powerpc/iommu: Add iommu_ops to report capabilities and allow blocking domains
authorAlexey Kardashevskiy <aik@ozlabs.ru>
Mon, 6 Mar 2023 17:31:00 +0000 (11:31 -0600)
committerMichael Ellerman <mpe@ellerman.id.au>
Tue, 14 Mar 2023 13:51:46 +0000 (00:51 +1100)
Up until now PPC64 managed to avoid using iommu_ops. The VFIO driver
uses a SPAPR TCE sub-driver and all iommu_ops uses were kept in the
Type1 VFIO driver. Recent development added 2 uses of iommu_ops to the
generic VFIO which broke POWER:
  - a coherency capability check;
  - blocking IOMMU domain - iommu_group_dma_owner_claimed()/...

This adds a simple iommu_ops which reports support for cache coherency
and provides a basic support for blocking domains. No other domain types
are implemented so the default domain is NULL.

Since now iommu_ops controls the group ownership, this takes it out of
VFIO.

This adds an IOMMU device into a pci_controller (=PHB) and registers it
in the IOMMU subsystem, iommu_ops is registered at this point. This
setup is done in postcore_initcall_sync.

This replaces iommu_group_add_device() with iommu_probe_device() as the
former misses necessary steps in connecting PCI devices to IOMMU
devices. This adds a comment about why explicit iommu_probe_device() is
still needed.

The previous discussion is here:
  https://lore.kernel.org/r/20220707135552.3688927-1-aik@ozlabs.ru/
  https://lore.kernel.org/r/20220701061751.1955857-1-aik@ozlabs.ru/

Fixes: e8ae0e140c05 ("vfio: Require that devices support DMA cache coherence")
Fixes: 70693f470848 ("vfio: Set DMA ownership for VFIO devices")
Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Timothy Pearson <tpearson@raptorengineering.com>
Acked-by: Alex Williamson <alex.williamson@redhat.com>
[mpe: Fix CONFIG_IOMMU_API=n build]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://msgid.link/2000135730.16998523.1678123860135.JavaMail.zimbra@raptorengineeringinc.com
arch/powerpc/include/asm/pci-bridge.h
arch/powerpc/kernel/iommu.c
arch/powerpc/platforms/powernv/pci-ioda.c
arch/powerpc/platforms/pseries/iommu.c
arch/powerpc/platforms/pseries/pseries.h
arch/powerpc/platforms/pseries/setup.c
drivers/vfio/vfio_iommu_spapr_tce.c

index 71c1d26..2aa3a09 100644 (file)
@@ -8,6 +8,7 @@
 #include <linux/list.h>
 #include <linux/ioport.h>
 #include <linux/numa.h>
+#include <linux/iommu.h>
 
 struct device_node;
 
@@ -44,6 +45,9 @@ struct pci_controller_ops {
 #endif
 
        void            (*shutdown)(struct pci_controller *hose);
+
+       struct iommu_group *(*device_group)(struct pci_controller *hose,
+                                           struct pci_dev *pdev);
 };
 
 /*
@@ -131,6 +135,9 @@ struct pci_controller {
        struct irq_domain       *dev_domain;
        struct irq_domain       *msi_domain;
        struct fwnode_handle    *fwnode;
+
+       /* iommu_ops support */
+       struct iommu_device     iommu;
 };
 
 /* These are used for config access before all the PCI probing
index 4cd3d68..0089dd4 100644 (file)
@@ -35,6 +35,7 @@
 #include <asm/vio.h>
 #include <asm/tce.h>
 #include <asm/mmu_context.h>
+#include <asm/ppc-pci.h>
 
 #define DBG(...)
 
@@ -1156,8 +1157,14 @@ int iommu_add_device(struct iommu_table_group *table_group, struct device *dev)
 
        pr_debug("%s: Adding %s to iommu group %d\n",
                 __func__, dev_name(dev),  iommu_group_id(table_group->group));
-
-       return iommu_group_add_device(table_group->group, dev);
+       /*
+        * This is still not adding devices via the IOMMU bus notifier because
+        * of pcibios_init() from arch/powerpc/kernel/pci_64.c which calls
+        * pcibios_scan_phb() first (and this guy adds devices and triggers
+        * the notifier) and only then it calls pci_bus_add_devices() which
+        * configures DMA for buses which also creates PEs and IOMMU groups.
+        */
+       return iommu_probe_device(dev);
 }
 EXPORT_SYMBOL_GPL(iommu_add_device);
 
@@ -1237,6 +1244,7 @@ static long spapr_tce_take_ownership(struct iommu_table_group *table_group)
                rc = iommu_take_ownership(tbl);
                if (!rc)
                        continue;
+
                for (j = 0; j < i; ++j)
                        iommu_release_ownership(table_group->tables[j]);
                return rc;
@@ -1269,4 +1277,140 @@ struct iommu_table_group_ops spapr_tce_table_group_ops = {
        .release_ownership = spapr_tce_release_ownership,
 };
 
+/*
+ * A simple iommu_ops to allow less cruft in generic VFIO code.
+ */
+static int spapr_tce_blocking_iommu_attach_dev(struct iommu_domain *dom,
+                                              struct device *dev)
+{
+       struct iommu_group *grp = iommu_group_get(dev);
+       struct iommu_table_group *table_group;
+       int ret = -EINVAL;
+
+       if (!grp)
+               return -ENODEV;
+
+       table_group = iommu_group_get_iommudata(grp);
+       ret = table_group->ops->take_ownership(table_group);
+       iommu_group_put(grp);
+
+       return ret;
+}
+
+static void spapr_tce_blocking_iommu_set_platform_dma(struct device *dev)
+{
+       struct iommu_group *grp = iommu_group_get(dev);
+       struct iommu_table_group *table_group;
+
+       table_group = iommu_group_get_iommudata(grp);
+       table_group->ops->release_ownership(table_group);
+}
+
+static const struct iommu_domain_ops spapr_tce_blocking_domain_ops = {
+       .attach_dev = spapr_tce_blocking_iommu_attach_dev,
+};
+
+static bool spapr_tce_iommu_capable(struct device *dev, enum iommu_cap cap)
+{
+       switch (cap) {
+       case IOMMU_CAP_CACHE_COHERENCY:
+               return true;
+       default:
+               break;
+       }
+
+       return false;
+}
+
+static struct iommu_domain *spapr_tce_iommu_domain_alloc(unsigned int type)
+{
+       struct iommu_domain *dom;
+
+       if (type != IOMMU_DOMAIN_BLOCKED)
+               return NULL;
+
+       dom = kzalloc(sizeof(*dom), GFP_KERNEL);
+       if (!dom)
+               return NULL;
+
+       dom->ops = &spapr_tce_blocking_domain_ops;
+
+       return dom;
+}
+
+static struct iommu_device *spapr_tce_iommu_probe_device(struct device *dev)
+{
+       struct pci_dev *pdev;
+       struct pci_controller *hose;
+
+       if (!dev_is_pci(dev))
+               return ERR_PTR(-EPERM);
+
+       pdev = to_pci_dev(dev);
+       hose = pdev->bus->sysdata;
+
+       return &hose->iommu;
+}
+
+static void spapr_tce_iommu_release_device(struct device *dev)
+{
+}
+
+static struct iommu_group *spapr_tce_iommu_device_group(struct device *dev)
+{
+       struct pci_controller *hose;
+       struct pci_dev *pdev;
+
+       pdev = to_pci_dev(dev);
+       hose = pdev->bus->sysdata;
+
+       if (!hose->controller_ops.device_group)
+               return ERR_PTR(-ENOENT);
+
+       return hose->controller_ops.device_group(hose, pdev);
+}
+
+static const struct iommu_ops spapr_tce_iommu_ops = {
+       .capable = spapr_tce_iommu_capable,
+       .domain_alloc = spapr_tce_iommu_domain_alloc,
+       .probe_device = spapr_tce_iommu_probe_device,
+       .release_device = spapr_tce_iommu_release_device,
+       .device_group = spapr_tce_iommu_device_group,
+       .set_platform_dma_ops = spapr_tce_blocking_iommu_set_platform_dma,
+};
+
+static struct attribute *spapr_tce_iommu_attrs[] = {
+       NULL,
+};
+
+static struct attribute_group spapr_tce_iommu_group = {
+       .name = "spapr-tce-iommu",
+       .attrs = spapr_tce_iommu_attrs,
+};
+
+static const struct attribute_group *spapr_tce_iommu_groups[] = {
+       &spapr_tce_iommu_group,
+       NULL,
+};
+
+/*
+ * This registers IOMMU devices of PHBs. This needs to happen
+ * after core_initcall(iommu_init) + postcore_initcall(pci_driver_init) and
+ * before subsys_initcall(iommu_subsys_init).
+ */
+static int __init spapr_tce_setup_phb_iommus_initcall(void)
+{
+       struct pci_controller *hose;
+
+       list_for_each_entry(hose, &hose_list, list_node) {
+               iommu_device_sysfs_add(&hose->iommu, hose->parent,
+                                      spapr_tce_iommu_groups, "iommu-phb%04x",
+                                      hose->global_number);
+               iommu_device_register(&hose->iommu, &spapr_tce_iommu_ops,
+                                     hose->parent);
+       }
+       return 0;
+}
+postcore_initcall_sync(spapr_tce_setup_phb_iommus_initcall);
+
 #endif /* CONFIG_IOMMU_API */
index c8a5cae..f851f49 100644 (file)
@@ -1899,6 +1899,13 @@ static long pnv_ioda2_take_ownership(struct iommu_table_group *table_group)
        /* Store @tbl as pnv_pci_ioda2_unset_window() resets it */
        struct iommu_table *tbl = pe->table_group.tables[0];
 
+       /*
+        * iommu_ops transfers the ownership per a device and we mode
+        * the group ownership with the first device in the group.
+        */
+       if (!tbl)
+               return 0;
+
        pnv_pci_ioda2_set_bypass(pe, false);
        pnv_pci_ioda2_unset_window(&pe->table_group, 0);
        if (pe->pbus)
@@ -1915,6 +1922,9 @@ static void pnv_ioda2_release_ownership(struct iommu_table_group *table_group)
        struct pnv_ioda_pe *pe = container_of(table_group, struct pnv_ioda_pe,
                                                table_group);
 
+       /* See the comment about iommu_ops above */
+       if (pe->table_group.tables[0])
+               return;
        pnv_pci_ioda2_setup_default_config(pe);
        if (pe->pbus)
                pnv_ioda_setup_bus_dma(pe, pe->pbus);
@@ -2921,6 +2931,27 @@ static void pnv_pci_ioda_dma_bus_setup(struct pci_bus *bus)
        }
 }
 
+#ifdef CONFIG_IOMMU_API
+static struct iommu_group *pnv_pci_device_group(struct pci_controller *hose,
+                                               struct pci_dev *pdev)
+{
+       struct pnv_phb *phb = hose->private_data;
+       struct pnv_ioda_pe *pe;
+
+       if (WARN_ON(!phb))
+               return ERR_PTR(-ENODEV);
+
+       pe = pnv_pci_bdfn_to_pe(phb, pdev->devfn | (pdev->bus->number << 8));
+       if (!pe)
+               return ERR_PTR(-ENODEV);
+
+       if (!pe->table_group.group)
+               return ERR_PTR(-ENODEV);
+
+       return iommu_group_ref_get(pe->table_group.group);
+}
+#endif
+
 static const struct pci_controller_ops pnv_pci_ioda_controller_ops = {
        .dma_dev_setup          = pnv_pci_ioda_dma_dev_setup,
        .dma_bus_setup          = pnv_pci_ioda_dma_bus_setup,
@@ -2931,6 +2962,9 @@ static const struct pci_controller_ops pnv_pci_ioda_controller_ops = {
        .setup_bridge           = pnv_pci_fixup_bridge_resources,
        .reset_secondary_bus    = pnv_pci_reset_secondary_bus,
        .shutdown               = pnv_pci_ioda_shutdown,
+#ifdef CONFIG_IOMMU_API
+       .device_group           = pnv_pci_device_group,
+#endif
 };
 
 static const struct pci_controller_ops pnv_npu_ocapi_ioda_controller_ops = {
index 8eec100..de77c8c 100644 (file)
@@ -1729,3 +1729,27 @@ static int __init tce_iommu_bus_notifier_init(void)
        return 0;
 }
 machine_subsys_initcall_sync(pseries, tce_iommu_bus_notifier_init);
+
+#ifdef CONFIG_SPAPR_TCE_IOMMU
+struct iommu_group *pSeries_pci_device_group(struct pci_controller *hose,
+                                            struct pci_dev *pdev)
+{
+       struct device_node *pdn, *dn = pdev->dev.of_node;
+       struct iommu_group *grp;
+       struct pci_dn *pci;
+
+       pdn = pci_dma_find(dn, NULL);
+       if (!pdn || !PCI_DN(pdn))
+               return ERR_PTR(-ENODEV);
+
+       pci = PCI_DN(pdn);
+       if (!pci->table_group)
+               return ERR_PTR(-ENODEV);
+
+       grp = pci->table_group->group;
+       if (!grp)
+               return ERR_PTR(-ENODEV);
+
+       return iommu_group_ref_get(grp);
+}
+#endif
index 1d75b77..f8bce40 100644 (file)
@@ -123,5 +123,9 @@ static inline void pseries_lpar_read_hblkrm_characteristics(void) { }
 #endif
 
 void pseries_rng_init(void);
+#ifdef CONFIG_SPAPR_TCE_IOMMU
+struct iommu_group *pSeries_pci_device_group(struct pci_controller *hose,
+                                            struct pci_dev *pdev);
+#endif
 
 #endif /* _PSERIES_PSERIES_H */
index 4a0cec8..94a7617 100644 (file)
@@ -1118,6 +1118,9 @@ static int pSeries_pci_probe_mode(struct pci_bus *bus)
 
 struct pci_controller_ops pseries_pci_controller_ops = {
        .probe_mode             = pSeries_pci_probe_mode,
+#ifdef CONFIG_SPAPR_TCE_IOMMU
+       .device_group           = pSeries_pci_device_group,
+#endif
 };
 
 define_machine(pseries) {
index c3f8ae1..a94ec62 100644 (file)
@@ -1200,8 +1200,6 @@ static void tce_iommu_release_ownership(struct tce_container *container,
        for (i = 0; i < IOMMU_TABLE_GROUP_MAX_TABLES; ++i)
                if (container->tables[i])
                        table_group->ops->unset_window(table_group, i);
-
-       table_group->ops->release_ownership(table_group);
 }
 
 static long tce_iommu_take_ownership(struct tce_container *container,
@@ -1209,10 +1207,6 @@ static long tce_iommu_take_ownership(struct tce_container *container,
 {
        long i, ret = 0;
 
-       ret = table_group->ops->take_ownership(table_group);
-       if (ret)
-               return ret;
-
        /* Set all windows to the new group */
        for (i = 0; i < IOMMU_TABLE_GROUP_MAX_TABLES; ++i) {
                struct iommu_table *tbl = container->tables[i];
@@ -1231,8 +1225,6 @@ 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;
 }