[Qemu-devel] [RFC v2 19/28] hw/vfio/common: Register specific nested mode notifiers and memory_listener

[Eric Auger]

In nested mode, legacy vfio_iommu_map_notify MAP/UNMAP notifier
cannot be used anymore. Indeed there is no caching mode in
place that allows to trap MAP events. Only configuration change
and UNMAP events can be trapped. As such we register
- one configuration notifier, whose role is to propagate the
  configuration update downto the host
- one UNMAP notifier, whose role is to propagate the TLB
  invalidation at physical IOMMU level.

Those notifiers propagate the guest stage 1 mappings at physical
level.

Also as there is no MAP event, the stage 2 mapping is not handled
anymore by the vfio_iommu_map_notify notifier.

We register a prereg_listener whose role is to dma_(un)map the RAM
memory regions. This programs the stage 2.

Signed-off-by: Eric Auger <address@hidden>

---

v1 -> v2:
- adapt to uapi changes
- pass the asid
- pass IOMMU_NOTIFIER_S1_CFG when initializing the config notifier
---
 hw/vfio/common.c | 167 ++++++++++++++++++++++++++++++++++++++---------
 1 file changed, 137 insertions(+), 30 deletions(-)

diff --git a/hw/vfio/common.c b/hw/vfio/common.c
index 83f5f2263d..77bdb76ade 100644
--- a/hw/vfio/common.c
+++ b/hw/vfio/common.c
@@ -349,6 +349,65 @@ static bool vfio_get_vaddr(IOMMUTLBEntry *iotlb, void 
**vaddr,
     return true;
 }
 
+/* Program the guest @cfg on physical IOMMU stage 1 (nested mode) */
+static void vfio_iommu_nested_notify(IOMMUNotifier *n,
+                                     IOMMUConfig *cfg)
+{
+    VFIOGuestIOMMU *giommu = container_of(n, VFIOGuestIOMMU, n);
+    VFIOContainer *container = giommu->container;
+    struct vfio_iommu_type1_bind_pasid_table info;
+    int ret;
+
+    info.argsz = sizeof(info);
+    info.flags = 0;
+    memcpy(&info.config, &cfg->pasid_cfg, sizeof(cfg->pasid_cfg));
+
+    ret = ioctl(container->fd, VFIO_IOMMU_BIND_PASID_TABLE, &info);
+    if (ret) {
+        error_report("%s: failed to pass S1 config to the host (%d)",
+                     __func__, ret);
+    }
+}
+
+/* Propagate a guest invalidation downto the physical IOMMU (nested mode) */
+static void vfio_iommu_unmap_notify(IOMMUNotifier *n, IOMMUTLBEntry *iotlb)
+{
+    VFIOGuestIOMMU *giommu = container_of(n, VFIOGuestIOMMU, n);
+    hwaddr start = iotlb->iova + giommu->iommu_offset;
+
+    VFIOContainer *container = giommu->container;
+    struct vfio_iommu_type1_cache_invalidate ustruct;
+    int ret;
+
+    assert(iotlb->perm == IOMMU_NONE);
+
+    ustruct.argsz = sizeof(ustruct);
+    ustruct.flags = 0;
+    ustruct.info.hdr.version = TLB_INV_HDR_VERSION_1;
+    ustruct.info.hdr.type = IOMMU_INV_TYPE_TLB;
+    ustruct.info.granularity = IOMMU_INV_NR_GRANU;
+    ustruct.info.flags = IOMMU_INVALIDATE_GLOBAL_PAGE;
+    /* 2^size of 4K pages, 0 for 4k, 9 for 2MB, etc. */
+    ustruct.info.size = ctz64(~iotlb->addr_mask) - 12;
+    /*
+     * TODO: at the moment we invalidate the whole ASID instead
+     * of invalidating the given nb_pages (nb_pages = 0):
+     * mask covering the whole GPA range is observed: in this case we shall
+     * invalidate the whole ASID (NH_ASID) and not induce storm of
+     * NH_VA commands.
+     */
+    ustruct.info.nr_pages = 0;
+    ustruct.info.addr = start;
+    ustruct.info.arch_id = iotlb->arch_id;
+
+    ret = ioctl(container->fd, VFIO_IOMMU_CACHE_INVALIDATE, &ustruct);
+    if (ret) {
+        error_report("%s: failed to invalidate CACHE for 0x%"PRIx64
+                     " mask=0x%"PRIx64" (%d)",
+                     __func__, start, iotlb->addr_mask, ret);
+    }
+}
+
 static void vfio_iommu_map_notify(IOMMUNotifier *n, IOMMUTLBEntry *iotlb)
 {
     VFIOGuestIOMMU *giommu = container_of(n, VFIOGuestIOMMU, n);
@@ -533,6 +592,32 @@ static void vfio_dma_unmap_ram_section(VFIOContainer 
*container,
     }
 }
 
+static void vfio_prereg_listener_region_add(MemoryListener *listener,
+                                            MemoryRegionSection *section)
+{
+    VFIOContainer *container =
+        container_of(listener, VFIOContainer, prereg_listener);
+
+    if (!memory_region_is_ram(section->mr)) {
+        return;
+    }
+
+    vfio_dma_map_ram_section(container, section);
+
+}
+static void vfio_prereg_listener_region_del(MemoryListener *listener,
+                                     MemoryRegionSection *section)
+{
+    VFIOContainer *container =
+        container_of(listener, VFIOContainer, prereg_listener);
+
+    if (!memory_region_is_ram(section->mr)) {
+        return;
+    }
+
+    vfio_dma_unmap_ram_section(container, section);
+}
+
 static void vfio_listener_region_add(MemoryListener *listener,
                                      MemoryRegionSection *section)
 {
@@ -541,7 +626,6 @@ static void vfio_listener_region_add(MemoryListener 
*listener,
     Int128 llend;
     int ret;
     VFIOHostDMAWindow *hostwin;
-    bool hostwin_found;
 
     if (vfio_listener_skipped_section(section)) {
         trace_vfio_listener_region_add_skip(
@@ -618,26 +702,10 @@ static void vfio_listener_region_add(MemoryListener 
*listener,
 #endif
     }
 
-    hostwin_found = false;
-    QLIST_FOREACH(hostwin, &container->hostwin_list, hostwin_next) {
-        if (hostwin->min_iova <= iova && end <= hostwin->max_iova) {
-            hostwin_found = true;
-            break;
-        }
-    }
-
-    if (!hostwin_found) {
-        error_report("vfio: IOMMU container %p can't map guest IOVA region"
-                     " 0x%"HWADDR_PRIx"..0x%"HWADDR_PRIx,
-                     container, iova, end);
-        ret = -EFAULT;
-        goto fail;
-    }
-
     memory_region_ref(section->mr);
 
     if (memory_region_is_iommu(section->mr)) {
-        VFIOGuestIOMMU *giommu;
+        VFIOGuestIOMMU *giommu = NULL;
         IOMMUMemoryRegion *iommu_mr = IOMMU_MEMORY_REGION(section->mr);
         hwaddr offset;
         int iommu_idx;
@@ -652,21 +720,40 @@ static void vfio_listener_region_add(MemoryListener 
*listener,
 
         offset = section->offset_within_address_space -
                     section->offset_within_region;
-        giommu = vfio_alloc_guest_iommu(container, iommu_mr, offset);
-
         llend = int128_add(int128_make64(section->offset_within_region),
                            section->size);
         llend = int128_sub(llend, int128_one());
         iommu_idx = memory_region_iommu_attrs_to_index(iommu_mr,
                                                        MEMTXATTRS_UNSPECIFIED);
-        iommu_iotlb_notifier_init(&giommu->n, vfio_iommu_map_notify,
-                                  IOMMU_NOTIFIER_IOTLB_ALL,
-                                  section->offset_within_region,
-                                  int128_get64(llend),
-                                  iommu_idx);
-        QLIST_INSERT_HEAD(&container->giommu_list, giommu, giommu_next);
 
-        memory_region_register_iommu_notifier(section->mr, &giommu->n);
+        if (container->iommu_type == VFIO_TYPE1_NESTING_IOMMU) {
+            /* Config notifier to propagate guest stage 1 config changes */
+            giommu = vfio_alloc_guest_iommu(container, iommu_mr, offset);
+            iommu_config_notifier_init(&giommu->n, vfio_iommu_nested_notify,
+                                       IOMMU_NOTIFIER_PASID_CFG, iommu_idx);
+            QLIST_INSERT_HEAD(&container->giommu_list, giommu, giommu_next);
+            memory_region_register_iommu_notifier(section->mr, &giommu->n);
+
+            /* IOTLB unmap notifier to propagate guest IOTLB invalidations */
+            giommu = vfio_alloc_guest_iommu(container, iommu_mr, offset);
+            iommu_iotlb_notifier_init(&giommu->n, vfio_iommu_unmap_notify,
+                                      IOMMU_NOTIFIER_UNMAP,
+                                      section->offset_within_region,
+                                      int128_get64(llend),
+                                      iommu_idx);
+            QLIST_INSERT_HEAD(&container->giommu_list, giommu, giommu_next);
+            memory_region_register_iommu_notifier(section->mr, &giommu->n);
+        } else {
+            /* MAP/UNMAP IOTLB notifier */
+            giommu = vfio_alloc_guest_iommu(container, iommu_mr, offset);
+            iommu_iotlb_notifier_init(&giommu->n, vfio_iommu_map_notify,
+                                      IOMMU_NOTIFIER_IOTLB_ALL,
+                                      section->offset_within_region,
+                                      int128_get64(llend),
+                                      iommu_idx);
+            QLIST_INSERT_HEAD(&container->giommu_list, giommu, giommu_next);
+            memory_region_register_iommu_notifier(section->mr, &giommu->n);
+        }
         memory_region_iommu_replay(giommu->iommu, &giommu->n);
 
         return;
@@ -679,7 +766,7 @@ static void vfio_listener_region_add(MemoryListener 
*listener,
     }
     return;
 
-fail:
+ fail:
     if (memory_region_is_ram_device(section->mr)) {
         error_report("failed to vfio_dma_map. pci p2p may not work");
         return;
@@ -763,15 +850,21 @@ static void vfio_listener_region_del(MemoryListener 
*listener,
     }
 }
 
-static const MemoryListener vfio_memory_listener = {
+static MemoryListener vfio_memory_listener = {
     .region_add = vfio_listener_region_add,
     .region_del = vfio_listener_region_del,
 };
 
+static MemoryListener vfio_memory_prereg_listener = {
+    .region_add = vfio_prereg_listener_region_add,
+    .region_del = vfio_prereg_listener_region_del,
+};
+
 static void vfio_listener_release(VFIOContainer *container)
 {
     memory_listener_unregister(&container->listener);
-    if (container->iommu_type == VFIO_SPAPR_TCE_v2_IOMMU) {
+    if (container->iommu_type == VFIO_SPAPR_TCE_v2_IOMMU ||
+        container->iommu_type == VFIO_TYPE1_NESTING_IOMMU) {
         memory_listener_unregister(&container->prereg_listener);
     }
 }
@@ -1262,6 +1355,20 @@ static int vfio_connect_container(VFIOGroup *group, 
AddressSpace *as,
         }
         vfio_host_win_add(container, 0, (hwaddr)-1, info.iova_pgsizes);
         container->pgsizes = info.iova_pgsizes;
+
+        if (container->iommu_type == VFIO_TYPE1_NESTING_IOMMU) {
+            container->prereg_listener = vfio_memory_prereg_listener;
+
+            memory_listener_register(&container->prereg_listener,
+                                     &address_space_memory);
+            if (container->error) {
+                memory_listener_unregister(&container->prereg_listener);
+                ret = container->error;
+                error_setg(errp,
+                    "RAM memory listener initialization failed for container");
+                goto free_container_exit;
+            }
+        }
         break;
     }
     case VFIO_SPAPR_TCE_v2_IOMMU:
-- 
2.17.1