[dpdk-dev] [PATCH 7/7] net/virtio: introduce Vhost-vDPA backend

Maxime Coquelin maxime.coquelin at redhat.com
Fri Sep 11 17:08:05 CEST 2020


vhost-vDPA is a new virtio backend type introduced by vDPA kernel
framework, which provides abstruction to the vDPA devices and
exposes an unified control interface through a char dev.

This patch adds support to the vhost-vDPA backend. As similar to
the existing vhost kernel backend, a set of virtio_user ops were
introduced for vhost-vDPA backend to handle device specific operations
such as:
 - device setup
 - ioctl message handling
 - queue pair enabling
 - dma map/unmap
vDPA relevant ioctl codes and data structures are also defined in
this patch.

Signed-off-by: Maxime Coquelin <maxime.coquelin at redhat.com>
---
 drivers/net/virtio/meson.build                |   1 +
 drivers/net/virtio/virtio_user/vhost.h        |   1 +
 drivers/net/virtio/virtio_user/vhost_vdpa.c   | 310 ++++++++++++++++++
 .../net/virtio/virtio_user/virtio_user_dev.c  |   9 +-
 4 files changed, 320 insertions(+), 1 deletion(-)
 create mode 100644 drivers/net/virtio/virtio_user/vhost_vdpa.c

diff --git a/drivers/net/virtio/meson.build b/drivers/net/virtio/meson.build
index 3fd6051f4b..eaed46373d 100644
--- a/drivers/net/virtio/meson.build
+++ b/drivers/net/virtio/meson.build
@@ -42,6 +42,7 @@ if is_linux
 		'virtio_user/vhost_kernel.c',
 		'virtio_user/vhost_kernel_tap.c',
 		'virtio_user/vhost_user.c',
+		'virtio_user/vhost_vdpa.c',
 		'virtio_user/virtio_user_dev.c')
 	deps += ['bus_vdev']
 endif
diff --git a/drivers/net/virtio/virtio_user/vhost.h b/drivers/net/virtio/virtio_user/vhost.h
index 2e71995a79..210a3704e7 100644
--- a/drivers/net/virtio/virtio_user/vhost.h
+++ b/drivers/net/virtio/virtio_user/vhost.h
@@ -113,5 +113,6 @@ struct virtio_user_backend_ops {
 
 extern struct virtio_user_backend_ops virtio_ops_user;
 extern struct virtio_user_backend_ops virtio_ops_kernel;
+extern struct virtio_user_backend_ops virtio_ops_vdpa;
 
 #endif
diff --git a/drivers/net/virtio/virtio_user/vhost_vdpa.c b/drivers/net/virtio/virtio_user/vhost_vdpa.c
new file mode 100644
index 0000000000..d959bca0a1
--- /dev/null
+++ b/drivers/net/virtio/virtio_user/vhost_vdpa.c
@@ -0,0 +1,310 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2020 Red Hat Inc.
+ */
+
+#include <sys/ioctl.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <fcntl.h>
+#include <unistd.h>
+
+#include <rte_memory.h>
+
+#include "vhost.h"
+#include "virtio_user_dev.h"
+
+struct vhost_memory_vdpa {
+	uint32_t nregions;
+	uint32_t padding;
+	struct vhost_memory_region regions[0];
+};
+
+/* vhost kernel & vdpa ioctls */
+#define VHOST_VIRTIO 0xAF
+#define VHOST_GET_FEATURES _IOR(VHOST_VIRTIO, 0x00, __u64)
+#define VHOST_SET_FEATURES _IOW(VHOST_VIRTIO, 0x00, __u64)
+#define VHOST_SET_OWNER _IO(VHOST_VIRTIO, 0x01)
+#define VHOST_RESET_OWNER _IO(VHOST_VIRTIO, 0x02)
+#define VHOST_SET_MEM_TABLE _IOW(VHOST_VIRTIO, 0x03, struct vhost_memory_vdpa)
+#define VHOST_SET_LOG_BASE _IOW(VHOST_VIRTIO, 0x04, __u64)
+#define VHOST_SET_LOG_FD _IOW(VHOST_VIRTIO, 0x07, int)
+#define VHOST_SET_VRING_NUM _IOW(VHOST_VIRTIO, 0x10, struct vhost_vring_state)
+#define VHOST_SET_VRING_ADDR _IOW(VHOST_VIRTIO, 0x11, struct vhost_vring_addr)
+#define VHOST_SET_VRING_BASE _IOW(VHOST_VIRTIO, 0x12, struct vhost_vring_state)
+#define VHOST_GET_VRING_BASE _IOWR(VHOST_VIRTIO, 0x12, struct vhost_vring_state)
+#define VHOST_SET_VRING_KICK _IOW(VHOST_VIRTIO, 0x20, struct vhost_vring_file)
+#define VHOST_SET_VRING_CALL _IOW(VHOST_VIRTIO, 0x21, struct vhost_vring_file)
+#define VHOST_SET_VRING_ERR _IOW(VHOST_VIRTIO, 0x22, struct vhost_vring_file)
+#define VHOST_NET_SET_BACKEND _IOW(VHOST_VIRTIO, 0x30, struct vhost_vring_file)
+#define VHOST_VDPA_GET_DEVICE_ID _IOR(VHOST_VIRTIO, 0x70, __u32)
+#define VHOST_VDPA_GET_STATUS _IOR(VHOST_VIRTIO, 0x71, __u8)
+#define VHOST_VDPA_SET_STATUS _IOW(VHOST_VIRTIO, 0x72, __u8)
+#define VHOST_VDPA_SET_VRING_ENABLE	_IOW(VHOST_VIRTIO, 0x75, \
+					     struct vhost_vring_state)
+
+static uint64_t vhost_req_user_to_vdpa[] = {
+	[VHOST_USER_SET_OWNER] = VHOST_SET_OWNER,
+	[VHOST_USER_RESET_OWNER] = VHOST_RESET_OWNER,
+	[VHOST_USER_SET_FEATURES] = VHOST_SET_FEATURES,
+	[VHOST_USER_GET_FEATURES] = VHOST_GET_FEATURES,
+	[VHOST_USER_SET_VRING_CALL] = VHOST_SET_VRING_CALL,
+	[VHOST_USER_SET_VRING_NUM] = VHOST_SET_VRING_NUM,
+	[VHOST_USER_SET_VRING_BASE] = VHOST_SET_VRING_BASE,
+	[VHOST_USER_GET_VRING_BASE] = VHOST_GET_VRING_BASE,
+	[VHOST_USER_SET_VRING_ADDR] = VHOST_SET_VRING_ADDR,
+	[VHOST_USER_SET_VRING_KICK] = VHOST_SET_VRING_KICK,
+	[VHOST_USER_SET_MEM_TABLE] = VHOST_SET_MEM_TABLE,
+	[VHOST_USER_SET_STATUS] = VHOST_VDPA_SET_STATUS,
+	[VHOST_USER_GET_STATUS] = VHOST_VDPA_GET_STATUS,
+	[VHOST_USER_SET_VRING_ENABLE] = VHOST_VDPA_SET_VRING_ENABLE,
+};
+
+/* no alignment requirement */
+struct vhost_iotlb_msg {
+	uint64_t iova;
+	uint64_t size;
+	uint64_t uaddr;
+#define VHOST_ACCESS_RO      0x1
+#define VHOST_ACCESS_WO      0x2
+#define VHOST_ACCESS_RW      0x3
+	uint8_t perm;
+#define VHOST_IOTLB_MISS           1
+#define VHOST_IOTLB_UPDATE         2
+#define VHOST_IOTLB_INVALIDATE     3
+#define VHOST_IOTLB_ACCESS_FAIL    4
+	uint8_t type;
+};
+
+#define VHOST_IOTLB_MSG_V2 0x2
+
+struct vhost_msg {
+	uint32_t type;
+	uint32_t reserved;
+	union {
+		struct vhost_iotlb_msg iotlb;
+		uint8_t padding[64];
+	};
+};
+
+static int
+vhost_vdpa_dma_map(struct virtio_user_dev *dev, void *addr,
+				  uint64_t iova, size_t len)
+{
+	struct vhost_msg msg = {};
+
+	msg.type = VHOST_IOTLB_MSG_V2;
+	msg.iotlb.type = VHOST_IOTLB_UPDATE;
+	msg.iotlb.iova = iova;
+	msg.iotlb.uaddr = (uint64_t)(uintptr_t)addr;
+	msg.iotlb.size = len;
+	msg.iotlb.perm = VHOST_ACCESS_RW;
+
+	if (write(dev->vhostfd, &msg, sizeof(msg)) != sizeof(msg)) {
+		PMD_DRV_LOG(ERR, "Failed to send IOTLB update (%s)",
+				strerror(errno));
+		return -1;
+	}
+
+	return 0;
+}
+
+static int
+vhost_vdpa_dma_unmap(struct virtio_user_dev *dev, void __rte_unused *addr,
+				  uint64_t iova, size_t len)
+{
+	struct vhost_msg msg = {};
+
+	msg.type = VHOST_IOTLB_MSG_V2;
+	msg.iotlb.type = VHOST_IOTLB_INVALIDATE;
+	msg.iotlb.iova = iova;
+	msg.iotlb.size = len;
+
+	if (write(dev->vhostfd, &msg, sizeof(msg)) != sizeof(msg)) {
+		PMD_DRV_LOG(ERR, "Failed to send IOTLB invalidate (%s)",
+				strerror(errno));
+		return -1;
+	}
+
+	return 0;
+}
+
+
+static int
+vhost_vdpa_map_contig(const struct rte_memseg_list *msl,
+		const struct rte_memseg *ms, size_t len, void *arg)
+{
+	struct virtio_user_dev *dev = arg;
+
+	if (msl->external)
+		return 0;
+
+	return vhost_vdpa_dma_map(dev, ms->addr, ms->iova, len);
+}
+
+static int
+vhost_vdpa_map(const struct rte_memseg_list *msl, const struct rte_memseg *ms,
+		void *arg)
+{
+	struct virtio_user_dev *dev = arg;
+
+	/* skip external memory that isn't a heap */
+	if (msl->external && !msl->heap)
+		return 0;
+
+	/* skip any segments with invalid IOVA addresses */
+	if (ms->iova == RTE_BAD_IOVA)
+		return 0;
+
+	/* if IOVA mode is VA, we've already mapped the internal segments */
+	if (!msl->external && rte_eal_iova_mode() == RTE_IOVA_VA)
+		return 0;
+
+	return vhost_vdpa_dma_map(dev, ms->addr, ms->iova, ms->len);
+}
+
+static int
+vhost_vdpa_dma_map_all(struct virtio_user_dev *dev)
+{
+	vhost_vdpa_dma_unmap(dev, 0, 0ULL, 0ULL - 1);
+
+	if (rte_eal_iova_mode() == RTE_IOVA_VA) {
+		/* with IOVA as VA mode, we can get away with mapping contiguous
+		 * chunks rather than going page-by-page.
+		 */
+		int ret = rte_memseg_contig_walk_thread_unsafe(
+				vhost_vdpa_map_contig, dev);
+		if (ret)
+			return ret;
+		/* we have to continue the walk because we've skipped the
+		 * external segments during the config walk.
+		 */
+	}
+	return rte_memseg_walk_thread_unsafe(vhost_vdpa_map, dev);
+}
+
+/* with below features, vhost vdpa does not need to do the checksum and TSO,
+ * these info will be passed to virtio_user through virtio net header.
+ */
+#define VHOST_VDPA_GUEST_OFFLOADS_MASK	\
+	((1ULL << VIRTIO_NET_F_GUEST_CSUM) |	\
+	 (1ULL << VIRTIO_NET_F_GUEST_TSO4) |	\
+	 (1ULL << VIRTIO_NET_F_GUEST_TSO6) |	\
+	 (1ULL << VIRTIO_NET_F_GUEST_ECN)  |	\
+	 (1ULL << VIRTIO_NET_F_GUEST_UFO))
+
+#define VHOST_VDPA_HOST_OFFLOADS_MASK		\
+	((1ULL << VIRTIO_NET_F_HOST_TSO4) |	\
+	 (1ULL << VIRTIO_NET_F_HOST_TSO6) |	\
+	 (1ULL << VIRTIO_NET_F_CSUM))
+
+static int
+vhost_vdpa_ioctl(struct virtio_user_dev *dev,
+		   enum vhost_user_request req,
+		   void *arg)
+{
+	int ret = -1;
+	uint64_t req_vdpa;
+	struct vhost_memory_vdpa *vm = NULL;
+
+	PMD_DRV_LOG(INFO, "%s", vhost_msg_strings[req]);
+
+	req_vdpa = vhost_req_user_to_vdpa[req];
+
+	if (req_vdpa == VHOST_SET_MEM_TABLE)
+		return vhost_vdpa_dma_map_all(dev);
+
+	if (req_vdpa == VHOST_SET_FEATURES) {
+		/* WORKAROUND */
+		*(uint64_t *)arg |= 1ULL << VIRTIO_F_IOMMU_PLATFORM;
+
+		/* Multiqueue not supported for now */
+		*(uint64_t *)arg &= ~(1ULL << VIRTIO_NET_F_MQ);
+	}
+
+	switch (req_vdpa) {
+	case VHOST_SET_VRING_NUM:
+	case VHOST_SET_VRING_ADDR:
+	case VHOST_SET_VRING_BASE:
+	case VHOST_GET_VRING_BASE:
+	case VHOST_SET_VRING_KICK:
+	case VHOST_SET_VRING_CALL:
+		*(unsigned int *)arg = *(unsigned int *)arg;
+		PMD_DRV_LOG(DEBUG, "vhostfd=%d, index=%u",
+			    dev->vhostfd, *(unsigned int *)arg);
+		break;
+	default:
+		break;
+	}
+
+	ret = ioctl(dev->vhostfd, req_vdpa, arg);
+
+	if (vm)
+		free(vm);
+
+	if (ret < 0)
+		PMD_DRV_LOG(ERR, "%s failed: %s",
+			    vhost_msg_strings[req], strerror(errno));
+
+	return ret;
+}
+
+/**
+ * Set up environment to talk with a vhost vdpa backend.
+ *
+ * @return
+ *   - (-1) if fail to set up;
+ *   - (>=0) if successful.
+ */
+static int
+vhost_vdpa_setup(struct virtio_user_dev *dev)
+{
+	uint32_t did = (uint32_t)-1;
+
+	dev->vhostfd = open(dev->path, O_RDWR);
+	if (dev->vhostfd < 0) {
+		PMD_DRV_LOG(ERR, "Failed to open %s: %s\n",
+				dev->path, strerror(errno));
+		return -1;
+	}
+
+	if (ioctl(dev->vhostfd, VHOST_VDPA_GET_DEVICE_ID, &did) < 0 ||
+			did != VIRTIO_ID_NETWORK) {
+		PMD_DRV_LOG(ERR, "Invalid vdpa device ID: %u\n", did);
+		return -1;
+	}
+
+	return 0;
+}
+
+static int
+vhost_vdpa_enable_queue_pair(struct virtio_user_dev *dev,
+			       uint16_t pair_idx,
+			       int enable)
+{
+	int i;
+
+	if (dev->qp_enabled[pair_idx] == enable)
+		return 0;
+
+	for (i = 0; i < 2; ++i) {
+		struct vhost_vring_state state = {
+			.index = pair_idx * 2 + i,
+			.num   = enable,
+		};
+
+		if (vhost_vdpa_ioctl(dev, VHOST_USER_SET_VRING_ENABLE, &state))
+			return -1;
+	}
+
+	dev->qp_enabled[pair_idx] = enable;
+
+	return 0;
+}
+
+struct virtio_user_backend_ops virtio_ops_vdpa = {
+	.setup = vhost_vdpa_setup,
+	.send_request = vhost_vdpa_ioctl,
+	.enable_qp = vhost_vdpa_enable_queue_pair,
+	.dma_map = vhost_vdpa_dma_map,
+	.dma_unmap = vhost_vdpa_dma_unmap,
+};
diff --git a/drivers/net/virtio/virtio_user/virtio_user_dev.c b/drivers/net/virtio/virtio_user/virtio_user_dev.c
index 2e097a95ea..2e8311147b 100644
--- a/drivers/net/virtio/virtio_user/virtio_user_dev.c
+++ b/drivers/net/virtio/virtio_user/virtio_user_dev.c
@@ -444,6 +444,12 @@ virtio_user_dev_setup(struct virtio_user_dev *dev)
 				dev->vhostfds[q] = -1;
 				dev->tapfds[q] = -1;
 			}
+		} else if (dev->backend_type ==
+				VIRTIO_USER_BACKEND_VHOST_VDPA) {
+			dev->ops = &virtio_ops_vdpa;
+		} else {
+			PMD_DRV_LOG(ERR, "Unknown backend type");
+			return -1;
 		}
 	}
 
@@ -878,7 +884,8 @@ virtio_user_update_status(struct virtio_user_dev *dev)
 	enum virtio_user_backend_type backend_type =
 				virtio_user_backend_type(dev->path);
 
-	if (backend_type != VIRTIO_USER_BACKEND_VHOST_USER)
+	if (backend_type != VIRTIO_USER_BACKEND_VHOST_USER &&
+			backend_type != VIRTIO_USER_BACKEND_VHOST_VDPA)
 		return 0;
 
 	err = dev->ops->send_request(dev, VHOST_USER_GET_STATUS, &ret);
-- 
2.26.2



More information about the dev mailing list