[PATCH v3 5/7] net/virtio_user: add vhost kernel support

[Jianfeng Tan <jianfeng.tan@intel.com>]

This patch add support vhost kernel as the backend for virtio_user.
Three main hook functions are added:
  - vhost_kernel_setup() to open char device, each vq pair needs one
    vhostfd;
  - vhost_kernel_ioctl() to communicate control messages with vhost
    kernel module;
  - vhost_kernel_enable_queue_pair() to open tap device and set it
    as the backend of corresonding vhost fd (that is to say, vq pair).

Signed-off-by: Jianfeng Tan <jianfeng.tan@intel.com>
---
 doc/guides/rel_notes/release_17_02.rst           |  20 ++
 drivers/net/virtio/Makefile                      |   1 +
 drivers/net/virtio/virtio_user/vhost.h           |   2 +
 drivers/net/virtio/virtio_user/vhost_kernel.c    | 373 +++++++++++++++++++++++
 drivers/net/virtio/virtio_user/virtio_user_dev.c |  21 +-
 drivers/net/virtio/virtio_user/virtio_user_dev.h |   6 +
 6 files changed, 420 insertions(+), 3 deletions(-)
 create mode 100644 drivers/net/virtio/virtio_user/vhost_kernel.c

diff --git a/doc/guides/rel_notes/release_17_02.rst b/doc/guides/rel_notes/release_17_02.rst
index 180af82..7354df5 100644
--- a/doc/guides/rel_notes/release_17_02.rst
+++ b/doc/guides/rel_notes/release_17_02.rst
@@ -52,6 +52,26 @@ New Features
   See the :ref:`Generic flow API <Generic_flow_API>` documentation for more
   information.
 
+* **virtio_user with vhost-kernel as another exceptional path.**
+
+  Previously, we upstreamed a virtual device, virtio_user with vhost-user
+  as the backend, as a way for IPC (Inter-Process Communication) and user
+  space container networking.
+
+  Virtio_user with vhost-kernel as the backend is a solution for exceptional
+  path, such as KNI, which exchanges packets with kernel networking stack.
+  This solution is very promising in:
+
+  * maintenance: vhost and vhost-net (kernel) is upstreamed and extensively
+    used kernel module.
+  * features: vhost-net is born to be a networking solution, which has
+    lots of networking related featuers, like multi queue, tso, multi-seg
+    mbuf, etc.
+  * performance: similar to KNI, this solution would use one or more
+    kthreads to send/receive packets from user space DPDK applications,
+    which has little impact on user space polling thread (except that
+    it might enter into kernel space to wake up those kthreads if
+    necessary).
 
 Resolved Issues
 ---------------
diff --git a/drivers/net/virtio/Makefile b/drivers/net/virtio/Makefile
index 97972a6..faeffb2 100644
--- a/drivers/net/virtio/Makefile
+++ b/drivers/net/virtio/Makefile
@@ -60,6 +60,7 @@ endif
 
 ifeq ($(CONFIG_RTE_VIRTIO_USER),y)
 SRCS-$(CONFIG_RTE_LIBRTE_VIRTIO_PMD) += virtio_user/vhost_user.c
+SRCS-$(CONFIG_RTE_LIBRTE_VIRTIO_PMD) += virtio_user/vhost_kernel.c
 SRCS-$(CONFIG_RTE_LIBRTE_VIRTIO_PMD) += virtio_user/virtio_user_dev.c
 SRCS-$(CONFIG_RTE_LIBRTE_VIRTIO_PMD) += virtio_user_ethdev.c
 endif
diff --git a/drivers/net/virtio/virtio_user/vhost.h b/drivers/net/virtio/virtio_user/vhost.h
index 515e4fc..5c983bd 100644
--- a/drivers/net/virtio/virtio_user/vhost.h
+++ b/drivers/net/virtio/virtio_user/vhost.h
@@ -118,4 +118,6 @@ struct virtio_user_backend_ops {
 };
 
 struct virtio_user_backend_ops ops_user;
+struct virtio_user_backend_ops ops_kernel;
+
 #endif
diff --git a/drivers/net/virtio/virtio_user/vhost_kernel.c b/drivers/net/virtio/virtio_user/vhost_kernel.c
new file mode 100644
index 0000000..1e7cdef
--- /dev/null
+++ b/drivers/net/virtio/virtio_user/vhost_kernel.c
@@ -0,0 +1,373 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2016 Intel Corporation. All rights reserved.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Intel Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <unistd.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <fcntl.h>
+#include <sys/ioctl.h>
+#include <net/if.h>
+#include <string.h>
+#include <errno.h>
+
+#include <rte_memory.h>
+#include <rte_eal_memconfig.h>
+
+#include "vhost.h"
+#include "virtio_user_dev.h"
+
+struct vhost_memory_kernel {
+	uint32_t nregions;
+	uint32_t padding;
+	struct vhost_memory_region regions[0];
+};
+
+/* vhost kernel 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_kernel)
+#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)
+
+/* TUN ioctls */
+#define TUNSETIFF     _IOW('T', 202, int)
+#define TUNGETFEATURES _IOR('T', 207, unsigned int)
+#define TUNSETOFFLOAD  _IOW('T', 208, unsigned int)
+#define TUNGETIFF      _IOR('T', 210, unsigned int)
+#define TUNSETSNDBUF   _IOW('T', 212, int)
+#define TUNGETVNETHDRSZ _IOR('T', 215, int)
+#define TUNSETVNETHDRSZ _IOW('T', 216, int)
+#define TUNSETQUEUE  _IOW('T', 217, int)
+#define TUNSETVNETLE _IOW('T', 220, int)
+#define TUNSETVNETBE _IOW('T', 222, int)
+
+/* TUNSETIFF ifr flags */
+#define IFF_TAP          0x0002
+#define IFF_NO_PI        0x1000
+#define IFF_ONE_QUEUE    0x2000
+#define IFF_VNET_HDR     0x4000
+#define IFF_MULTI_QUEUE  0x0100
+#define IFF_ATTACH_QUEUE 0x0200
+#define IFF_DETACH_QUEUE 0x0400
+
+/* Constants */
+#define TUN_DEF_SNDBUF	(1ull << 20)
+#define PATH_NET_TUN	"/dev/net/tun"
+#define VHOST_KERNEL_MAX_REGIONS	64
+
+static uint64_t vhost_req_user_to_kernel[] = {
+	[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,
+};
+
+/* By default, vhost kernel module allows 64 regions, but DPDK allows
+ * 256 segments. As a relief, below function merges those virtually
+ * adjacent memsegs into one region.
+ */
+static struct vhost_memory_kernel *
+prepare_vhost_memory_kernel(void)
+{
+	uint32_t i, j, k = 0;
+	struct rte_memseg *seg;
+	struct vhost_memory_region *mr;
+	struct vhost_memory_kernel *vm;
+
+	vm = malloc(sizeof(struct vhost_memory_kernel) +
+		    VHOST_KERNEL_MAX_REGIONS *
+		    sizeof(struct vhost_memory_region));
+
+	for (i = 0; i < RTE_MAX_MEMSEG; ++i) {
+		seg = &rte_eal_get_configuration()->mem_config->memseg[i];
+		if (!seg->addr)
+			break;
+
+		int new_region = 1;
+
+		for (j = 0; j < k; ++j) {
+			mr = &vm->regions[j];
+
+			if (mr->userspace_addr + mr->memory_size ==
+			    (uint64_t)(uintptr_t)seg->addr) {
+				mr->memory_size += seg->len;
+				new_region = 0;
+				break;
+			}
+
+			if ((uint64_t)(uintptr_t)seg->addr + seg->len ==
+			    mr->userspace_addr) {
+				mr->guest_phys_addr =
+					(uint64_t)(uintptr_t)seg->addr;
+				mr->userspace_addr =
+					(uint64_t)(uintptr_t)seg->addr;
+				mr->memory_size += seg->len;
+				new_region = 0;
+				break;
+			}
+		}
+
+		if (new_region == 0)
+			continue;
+
+		mr = &vm->regions[k++];
+		/* use vaddr here! */
+		mr->guest_phys_addr = (uint64_t)(uintptr_t)seg->addr;
+		mr->userspace_addr = (uint64_t)(uintptr_t)seg->addr;
+		mr->memory_size = seg->len;
+		mr->mmap_offset = 0;
+
+		if (k >= VHOST_KERNEL_MAX_REGIONS) {
+			free(vm);
+			return NULL;
+		}
+	}
+
+	vm->nregions = k;
+	vm->padding = 0;
+	return vm;
+}
+
+static int
+vhost_kernel_ioctl(struct virtio_user_dev *dev,
+		   enum vhost_user_request req,
+		   void *arg)
+{
+	int i, ret = -1;
+	uint64_t req_kernel;
+	struct vhost_memory_kernel *vm = NULL;
+
+	PMD_DRV_LOG(INFO, "%s", vhost_msg_strings[req]);
+
+	req_kernel = vhost_req_user_to_kernel[req];
+
+	if (req_kernel == VHOST_SET_MEM_TABLE) {
+		vm = prepare_vhost_memory_kernel();
+		if (!vm)
+			return -1;
+		arg = (void *)vm;
+	}
+
+	/* Does not work when VIRTIO_F_IOMMU_PLATFORM now, why? */
+	if (req_kernel == VHOST_SET_FEATURES)
+		*(uint64_t *)arg &= ~(1ULL << VIRTIO_F_IOMMU_PLATFORM);
+
+	for (i = 0; i < VHOST_KERNEL_MAX_QUEUES; ++i) {
+		if (dev->vhostfds[i] < 0)
+			continue;
+
+		ret = ioctl(dev->vhostfds[i], req_kernel, arg);
+		if (ret < 0)
+			break;
+	}
+
+	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 kernel backend.
+ *
+ * @return
+ *   - (-1) if fail to set up;
+ *   - (>=0) if successful.
+ */
+static int
+vhost_kernel_setup(struct virtio_user_dev *dev)
+{
+	int vhostfd;
+	uint32_t i;
+
+	for (i = 0; i < dev->max_queue_pairs; ++i) {
+		vhostfd = open(dev->path, O_RDWR);
+		if (vhostfd < 0) {
+			PMD_DRV_LOG(ERR, "fail to open %s, %s",
+				    dev->path, strerror(errno));
+			return -1;
+		}
+
+		dev->vhostfds[i] = vhostfd;
+	}
+
+	return 0;
+}
+
+static int
+vhost_kernel_set_backend(int vhostfd, int tapfd)
+{
+	struct vhost_vring_file f;
+
+	f.fd = tapfd;
+	f.index = 0;
+	if (ioctl(vhostfd, VHOST_NET_SET_BACKEND, &f) < 0) {
+		PMD_DRV_LOG(ERR, "VHOST_NET_SET_BACKEND fails, %s",
+				strerror(errno));
+		return -1;
+	}
+
+	f.index = 1;
+	if (ioctl(vhostfd, VHOST_NET_SET_BACKEND, &f) < 0) {
+		PMD_DRV_LOG(ERR, "VHOST_NET_SET_BACKEND fails, %s",
+				strerror(errno));
+		return -1;
+	}
+
+	return 0;
+}
+
+static int
+vhost_kernel_enable_queue_pair(struct virtio_user_dev *dev,
+			       uint16_t pair_idx,
+			       int enable)
+{
+	unsigned int tap_features;
+	int sndbuf = TUN_DEF_SNDBUF;
+	struct ifreq ifr;
+	int hdr_size;
+	int vhostfd;
+	int tapfd;
+
+	vhostfd = dev->vhostfds[pair_idx];
+
+	if (!enable) {
+		if (dev->tapfds[pair_idx]) {
+			close(dev->tapfds[pair_idx]);
+			dev->tapfds[pair_idx] = -1;
+		}
+		return vhost_kernel_set_backend(vhostfd, -1);
+	} else if (dev->tapfds[pair_idx] >= 0) {
+		return 0;
+	}
+
+	if ((dev->features & (1ULL << VIRTIO_NET_F_MRG_RXBUF)) ||
+	    (dev->features & (1ULL << VIRTIO_F_VERSION_1)))
+		hdr_size = sizeof(struct virtio_net_hdr_mrg_rxbuf);
+	else
+		hdr_size = sizeof(struct virtio_net_hdr);
+
+	/* TODO:
+	 * 1. verify we can get/set vnet_hdr_len, tap_probe_vnet_hdr_len
+	 * 2. get number of memory regions from vhost module parameter
+	 * max_mem_regions, supported in newer version linux kernel
+	 */
+	tapfd = open(PATH_NET_TUN, O_RDWR);
+	if (tapfd < 0) {
+		PMD_DRV_LOG(ERR, "fail to open %s: %s",
+			    PATH_NET_TUN, strerror(errno));
+		return -1;
+	}
+
+	/* Construct ifr */
+	memset(&ifr, 0, sizeof(ifr));
+	ifr.ifr_flags = IFF_TAP | IFF_NO_PI;
+
+	if (ioctl(tapfd, TUNGETFEATURES, &tap_features) == -1) {
+		PMD_DRV_LOG(ERR, "TUNGETFEATURES failed: %s", strerror(errno));
+		goto error;
+	}
+	if (tap_features & IFF_ONE_QUEUE)
+		ifr.ifr_flags |= IFF_ONE_QUEUE;
+
+	/* Let tap instead of vhost-net handle vnet header, as the latter does
+	 * not support offloading. And in this case, we should not set feature
+	 * bit VHOST_NET_F_VIRTIO_NET_HDR.
+	 */
+	if (tap_features & IFF_VNET_HDR) {
+		ifr.ifr_flags |= IFF_VNET_HDR;
+	} else {
+		PMD_DRV_LOG(ERR, "TAP does not support IFF_VNET_HDR");
+		goto error;
+	}
+
+	if (dev->ifname)
+		strncpy(ifr.ifr_name, dev->ifname, IFNAMSIZ);
+	else
+		strncpy(ifr.ifr_name, "tap%d", IFNAMSIZ);
+	if (ioctl(tapfd, TUNSETIFF, (void *)&ifr) == -1) {
+		PMD_DRV_LOG(ERR, "TUNSETIFF failed: %s", strerror(errno));
+		goto error;
+	}
+
+	fcntl(tapfd, F_SETFL, O_NONBLOCK);
+
+	if (ioctl(tapfd, TUNSETVNETHDRSZ, &hdr_size) < 0) {
+		PMD_DRV_LOG(ERR, "TUNSETVNETHDRSZ failed: %s", strerror(errno));
+		goto error;
+	}
+
+	if (ioctl(tapfd, TUNSETSNDBUF, &sndbuf) < 0) {
+		PMD_DRV_LOG(ERR, "TUNSETSNDBUF failed: %s", strerror(errno));
+		goto error;
+	}
+
+	if (vhost_kernel_set_backend(vhostfd, tapfd) < 0)
+		goto error;
+
+	dev->tapfds[pair_idx] = tapfd;
+	if (!dev->ifname)
+		dev->ifname = strdup(ifr.ifr_name);
+
+	return 0;
+error:
+	return -1;
+}
+
+struct virtio_user_backend_ops ops_kernel = {
+	.setup = vhost_kernel_setup,
+	.send_request = vhost_kernel_ioctl,
+	.enable_qp = vhost_kernel_enable_queue_pair
+};
diff --git a/drivers/net/virtio/virtio_user/virtio_user_dev.c b/drivers/net/virtio/virtio_user/virtio_user_dev.c
index 32039a1..c40b77e 100644
--- a/drivers/net/virtio/virtio_user/virtio_user_dev.c
+++ b/drivers/net/virtio/virtio_user/virtio_user_dev.c
@@ -192,6 +192,9 @@ int virtio_user_stop_device(struct virtio_user_dev *dev)
 	for (i = 0; i < dev->max_queue_pairs; ++i)
 		dev->ops->enable_qp(dev, i, 0);
 
+	free(dev->ifname);
+	dev->ifname = NULL;
+
 	return 0;
 }
 
@@ -230,7 +233,7 @@ is_vhost_user_by_type(const char *path)
 static int
 virtio_user_dev_setup(struct virtio_user_dev *dev)
 {
-	uint32_t i;
+	uint32_t i, q;
 
 	dev->vhostfd = -1;
 	for (i = 0; i < VIRTIO_MAX_VIRTQUEUES * 2 + 1; ++i) {
@@ -238,12 +241,18 @@ virtio_user_dev_setup(struct virtio_user_dev *dev)
 		dev->callfds[i] = -1;
 	}
 
+	for (q = 0; q < VHOST_KERNEL_MAX_QUEUES; ++q) {
+		dev->vhostfds[q] = -1;
+		dev->tapfds[q] = -1;
+	}
+
 	if (is_vhost_user_by_type(dev->path)) {
 		dev->ops = &ops_user;
-		return dev->ops->setup(dev);
+	} else {
+		dev->ops = &ops_kernel;
 	}
 
-	return -1;
+	return dev->ops->setup(dev);
 }
 
 int
@@ -295,7 +304,13 @@ virtio_user_dev_init(struct virtio_user_dev *dev, char *path, int queues,
 void
 virtio_user_dev_uninit(struct virtio_user_dev *dev)
 {
+	uint32_t i;
+
+	virtio_user_stop_device(dev);
+
 	close(dev->vhostfd);
+	for (i = 0; i < VHOST_KERNEL_MAX_QUEUES; ++i)
+		close(dev->vhostfds[i]);
 }
 
 static uint8_t
diff --git a/drivers/net/virtio/virtio_user/virtio_user_dev.h b/drivers/net/virtio/virtio_user/virtio_user_dev.h
index 9f2f82e..148b2e6 100644
--- a/drivers/net/virtio/virtio_user/virtio_user_dev.h
+++ b/drivers/net/virtio/virtio_user/virtio_user_dev.h
@@ -43,6 +43,12 @@ struct virtio_user_dev {
 	/* for vhost_user backend */
 	int		vhostfd;
 
+	/* for vhost_kernel backend */
+	char		*ifname;
+#define VHOST_KERNEL_MAX_QUEUES		8
+	int		vhostfds[VHOST_KERNEL_MAX_QUEUES];
+	int		tapfds[VHOST_KERNEL_MAX_QUEUES];
+
 	/* for both vhost_user and vhost_kernel */
 	int		callfds[VIRTIO_MAX_VIRTQUEUES * 2 + 1];
 	int		kickfds[VIRTIO_MAX_VIRTQUEUES * 2 + 1];
-- 
2.7.4