qemu/vdpa-iommufd-Implement-DMA-mapping-through-the-iommu.patch

From b88b03c84aa695b96a91329e2d01fffad551c34d Mon Sep 17 00:00:00 2001
From: libai <libai12@huawei.com>
Date: Thu, 27 Mar 2025 19:24:53 +0800
Subject: [PATCH] vdpa/iommufd:Implement DMA mapping through the iommufd
 interface

Change the owner of memorylistener from the independent vDPA device to VDPAIOMMUFDContainer

Signed-off-by: libai <libai12@huawei.com>
---
 hw/virtio/vdpa-dev-iommufd.c         | 137 +++++++++++++++++++++++++++
 hw/virtio/vdpa-dev.c                 |   4 +-
 hw/virtio/vhost-vdpa.c               |  13 +--
 include/hw/virtio/vdpa-dev-iommufd.h |   1 +
 include/hw/virtio/vhost-vdpa.h       |   7 ++
 5 files changed, 154 insertions(+), 8 deletions(-)

diff --git a/hw/virtio/vdpa-dev-iommufd.c b/hw/virtio/vdpa-dev-iommufd.c
index d72f56d52f..668c6a1cb1 100644
--- a/hw/virtio/vdpa-dev-iommufd.c
+++ b/hw/virtio/vdpa-dev-iommufd.c
@@ -9,11 +9,124 @@
 #include <sys/ioctl.h>
 #include <linux/vhost.h>
 #include "qapi/error.h"
+#include "exec/target_page.h"
+#include "exec/address-spaces.h"
 #include "hw/virtio/vdpa-dev-iommufd.h"
 
 static QLIST_HEAD(, VDPAIOMMUFDContainer) vdpa_container_list =
     QLIST_HEAD_INITIALIZER(vdpa_container_list);
 
+static int vhost_vdpa_iommufd_container_dma_map(VDPAIOMMUFDContainer *container, hwaddr iova,
+                                                hwaddr size, void *vaddr, bool readonly)
+{
+    return iommufd_backend_map_dma(container->iommufd, container->ioas_id, iova, size, vaddr, readonly);
+
+}
+static int vhost_vdpa_iommufd_container_dma_unmap(VDPAIOMMUFDContainer *container,
+                                                  hwaddr iova, hwaddr size)
+{
+    return iommufd_backend_unmap_dma(container->iommufd, container->ioas_id, iova, size);
+}
+
+static void vhost_vdpa_iommufd_container_region_add(MemoryListener *listener,
+                                                    MemoryRegionSection *section)
+{
+    VDPAIOMMUFDContainer *container = container_of(listener, VDPAIOMMUFDContainer, listener);
+    hwaddr iova;
+    Int128 llend, llsize;
+    void *vaddr;
+    int page_size = qemu_target_page_size();
+    int page_mask = -page_size;
+    int ret;
+
+    if (vhost_vdpa_listener_skipped_section(section, 0, ULLONG_MAX, page_mask)) {
+        return;
+    }
+
+    if (unlikely((section->offset_within_address_space & ~page_mask) !=
+                 (section->offset_within_region & ~page_mask))) {
+        return;
+    }
+
+    iova = ROUND_UP(section->offset_within_address_space, page_size);
+    llend = vhost_vdpa_section_end(section, page_mask);
+    if (int128_ge(int128_make64(iova), llend)) {
+        return;
+    }
+
+    memory_region_ref(section->mr);
+    vaddr = memory_region_get_ram_ptr(section->mr) +
+            section->offset_within_region +
+            (iova - section->offset_within_address_space);
+
+    llsize = int128_sub(llend, int128_make64(iova));
+
+    ret = vhost_vdpa_iommufd_container_dma_map(container, iova, int128_get64(llsize),
+                                               vaddr, section->readonly);
+    if (ret) {
+        qemu_log("vhost vdpa iommufd container dma map failed: %d\n", ret);
+    }
+}
+
+static void vhost_vdpa_iommufd_container_region_del(MemoryListener *listener,
+                                                    MemoryRegionSection *section)
+{
+    VDPAIOMMUFDContainer *container = container_of(listener, VDPAIOMMUFDContainer, listener);
+    hwaddr iova;
+    Int128 llend, llsize;
+    int page_size = qemu_target_page_size();
+    int page_mask = -page_size;
+    int ret;
+
+    if (vhost_vdpa_listener_skipped_section(section, 0, ULLONG_MAX, page_mask)) {
+        return;
+    }
+
+    if (unlikely((section->offset_within_address_space & ~page_mask) !=
+                 (section->offset_within_region & ~page_mask))) {
+        return;
+    }
+
+    iova = ROUND_UP(section->offset_within_address_space, page_size);
+    llend = vhost_vdpa_section_end(section, page_mask);
+
+    if (int128_ge(int128_make64(iova), llend)) {
+        return;
+    }
+
+    llsize = int128_sub(llend, int128_make64(iova));
+    /*
+     * The unmap ioctl doesn't accept a full 64-bit. need to check it
+     */
+    if (int128_eq(llsize, int128_2_64())) {
+        llsize = int128_rshift(llsize, 1);
+        ret = vhost_vdpa_iommufd_container_dma_unmap(container, iova, int128_get64(llsize));
+
+        if (ret) {
+            qemu_log("vhost vdpa iommufd container unmap failed(0x%" HWADDR_PRIx ", "
+                     "0x%" HWADDR_PRIx ") = %d (%m)", iova, int128_get64(llsize), ret);
+        }
+        iova += int128_get64(llsize);
+    }
+    ret = vhost_vdpa_iommufd_container_dma_unmap(container, iova, int128_get64(llsize));
+
+    if (ret) {
+        qemu_log("vhost vdpa iommufd container unmap failed(0x%" HWADDR_PRIx ", "
+                  "0x%" HWADDR_PRIx ") = %d (%m)", iova, int128_get64(llsize), ret);
+    }
+
+    memory_region_unref(section->mr);
+}
+
+/*
+ * IOTLB API used by vhost vdpa iommufd container
+ */
+const MemoryListener vhost_vdpa_iommufd_container_listener = {
+    .name = "vhost-vdpa-iommufd-container",
+    .region_add = vhost_vdpa_iommufd_container_region_add,
+    .region_del = vhost_vdpa_iommufd_container_region_del,
+};
+
 static int vhost_vdpa_container_connect_iommufd(VDPAIOMMUFDContainer *container)
 {
     IOMMUFDBackend *iommufd = container->iommufd;
@@ -87,6 +200,7 @@ static VDPAIOMMUFDContainer *vhost_vdpa_create_container(VhostVdpaDevice *vdev)
 
     container = g_new0(VDPAIOMMUFDContainer, 1);
     container->iommufd = vdev->iommufd;
+    container->listener = vhost_vdpa_iommufd_container_listener;
     QLIST_INIT(&container->hwpt_list);
 
     QLIST_INSERT_HEAD(&vdpa_container_list, container, next);
@@ -213,11 +327,27 @@ static void vhost_vdpa_container_detach_device(VDPAIOMMUFDContainer *container,
     }
 }
 
+static int vhost_vdpa_container_get_dev_count(VDPAIOMMUFDContainer *container)
+{
+    IOMMUFDHWPT *hwpt;
+    VhostVdpaDevice *dev;
+    int dev_count = 0;
+
+    QLIST_FOREACH(hwpt, &container->hwpt_list, next) {
+        QLIST_FOREACH(dev, &hwpt->device_list, next) {
+            dev_count++;
+        }
+    }
+
+    return dev_count;
+}
+
 int vhost_vdpa_attach_container(VhostVdpaDevice *vdev)
 {
     VDPAIOMMUFDContainer *container = NULL;
     IOMMUFDBackend *iommufd = vdev->iommufd;
     bool new_container = false;
+    int dev_count = 0;
     int ret = 0;
 
     if (!iommufd) {
@@ -251,6 +381,12 @@ int vhost_vdpa_attach_container(VhostVdpaDevice *vdev)
         goto unbind;
     }
 
+    /* register the container memory listener when attaching the first device */
+    dev_count = vhost_vdpa_container_get_dev_count(container);
+    if (dev_count == 1) {
+        memory_listener_register(&container->listener, &address_space_memory);
+    }
+
     return 0;
 
 unbind:
@@ -288,6 +424,7 @@ void vhost_vdpa_detach_container(VhostVdpaDevice *vdev)
         return;
     }
     /* No HWPT in this container, destroy it */
+    memory_listener_unregister(&container->listener);
     vhost_vdpa_container_disconnect_iommufd(container);
 
     vhost_vdpa_destroy_container(container);
diff --git a/hw/virtio/vdpa-dev.c b/hw/virtio/vdpa-dev.c
index a6bd695724..b256ad540c 100644
--- a/hw/virtio/vdpa-dev.c
+++ b/hw/virtio/vdpa-dev.c
@@ -136,9 +136,9 @@ static void vhost_vdpa_device_realize(DeviceState *dev, Error **errp)
                        strerror(-ret));
             goto free_vqs;
         }
+    } else {
+        memory_listener_register(&v->vdpa.listener, &address_space_memory);
     }
-
-    memory_listener_register(&v->vdpa.listener, &address_space_memory);
     v->config_size = vhost_vdpa_device_get_u32(v->vhostfd,
                                                VHOST_VDPA_GET_CONFIG_SIZE,
                                                errp);
diff --git a/hw/virtio/vhost-vdpa.c b/hw/virtio/vhost-vdpa.c
index 4a8fc37851..b5fb89b98e 100644
--- a/hw/virtio/vhost-vdpa.c
+++ b/hw/virtio/vhost-vdpa.c
@@ -26,13 +26,14 @@
 #include "qemu/main-loop.h"
 #include "trace.h"
 #include "qapi/error.h"
+#include "hw/virtio/vdpa-dev-iommufd.h"
 
 /*
  * Return one past the end of the end of section. Be careful with uint64_t
  * conversions!
  */
-static Int128 vhost_vdpa_section_end(const MemoryRegionSection *section,
-                                     int page_mask)
+Int128 vhost_vdpa_section_end(const MemoryRegionSection *section,
+                              int page_mask)
 {
     Int128 llend = int128_make64(section->offset_within_address_space);
     llend = int128_add(llend, section->size);
@@ -41,10 +42,10 @@ static Int128 vhost_vdpa_section_end(const MemoryRegionSection *section,
     return llend;
 }
 
-static bool vhost_vdpa_listener_skipped_section(MemoryRegionSection *section,
-                                                uint64_t iova_min,
-                                                uint64_t iova_max,
-                                                int page_mask)
+bool vhost_vdpa_listener_skipped_section(MemoryRegionSection *section,
+                                         uint64_t iova_min,
+                                         uint64_t iova_max,
+                                         int page_mask)
 {
     Int128 llend;
 
diff --git a/include/hw/virtio/vdpa-dev-iommufd.h b/include/hw/virtio/vdpa-dev-iommufd.h
index dc14d9dd15..8e56647690 100644
--- a/include/hw/virtio/vdpa-dev-iommufd.h
+++ b/include/hw/virtio/vdpa-dev-iommufd.h
@@ -23,6 +23,7 @@ typedef struct IOMMUFDHWPT {
 } IOMMUFDHWPT;
 
 typedef struct VDPAIOMMUFDContainer {
+    MemoryListener listener;
     struct IOMMUFDBackend *iommufd;
     uint32_t ioas_id;
     QLIST_HEAD(, IOMMUFDHWPT) hwpt_list;
diff --git a/include/hw/virtio/vhost-vdpa.h b/include/hw/virtio/vhost-vdpa.h
index ee255bc1bd..e32effc6e1 100644
--- a/include/hw/virtio/vhost-vdpa.h
+++ b/include/hw/virtio/vhost-vdpa.h
@@ -57,6 +57,13 @@ typedef struct vhost_vdpa {
 int vhost_vdpa_get_iova_range(int fd, struct vhost_vdpa_iova_range *iova_range);
 int vhost_vdpa_set_vring_ready(struct vhost_vdpa *v, unsigned idx);
 
+Int128 vhost_vdpa_section_end(const MemoryRegionSection *section,
+                              int page_mask);
+bool vhost_vdpa_listener_skipped_section(MemoryRegionSection *section,
+                                         uint64_t iova_min,
+                                         uint64_t iova_max,
+                                         int page_mask);
+
 int vhost_vdpa_dma_map(struct vhost_vdpa *v, uint32_t asid, hwaddr iova,
                        hwaddr size, void *vaddr, bool readonly);
 int vhost_vdpa_dma_unmap(struct vhost_vdpa *v, uint32_t asid, hwaddr iova,
-- 
2.41.0.windows.1