总述：OVS DPDK与QEMU之间如何通过vhost user协议通信
vhost user协议的控制和数据通道
所有的控制信息通过UNIX套接口（控制通道）交互。包括为进行直接内存访问而交换的内存映射信息，以及当数据填入virtio队列后需要出发的kick事件和中断信息。在Neutron中此UNIX套接口命名为vhuxxxxxxxx-xx；
数据通道事实上由内存直接访问实现。客户机中的virtio-net驱动分配一部分内存用于virtio的队列。virtio标准定义了此队列的结构。QEMU通过控制通道将此部分内存的地址共享给OVS DPDK。DPDK自身映射一个相同标准的virtio队列结构到此内存上，藉此来读写客户机巨页内存中的virtio队列。直接内存访问的实现需要在OVS DPDK和QEMU之间使用巨页内存。如果QEMU设置正确，但是没有配置巨页内存，OVS DPDK将不能访问QEMU的内存，二者也就不能交换数据报文。如果用户忘记了请求客户机巨页内存，nova将通过宏数据通知用户。
当OVS DPDK向客户机发送数据包时，这些数据包在OVS DPDK的统计里面显示为接口vhuxxxxxxxx-xx的发送Tx流量。在客户机中，显示为接收Rx流量。
当客户机向OVS DPDK发送数据包时，这些数据包在客户机中显示为发送Tx流量，而在OVS DPDK中显示为接口vhuxxxxxxxx-xx的接收Rx流量。
客户机并没有硬件的统计计数。ethtool工具的-s选项未实现。所有的底层统计计数只能使用OVS的命令显示（ovs-vsctl list get interfave vhuxxxxxxxx-xx statistics），因此显示的数据都是基于OVS DPDK的视角。
虽然数据包可通过共享内存传输，但是还需要一种方法告知对端数据包已经拷贝到virtio队列中。通过vhost user套接口vhuxxxxxxxx-xx实现的控制通道可用来完成通知（kicking）对方的功能。通知必然有代价。首先，需要一个写套接口的系统调用；之后对端需要处理一个中断操作。所以，接收双方都会在控制通道上消耗时间。
为避免控制通道的通知消耗，OpenvSwitch和QEMU都可以设置特殊标志以告知对方其不愿接收中断。尽管如此，只有在采用临时或者固定查询virtio队列方式时才能使用不接收中断的功能。
为客户机的性能考虑其本身可采用DPDK处理数据包。尽管Linux内核采用轮询处理和中断相结合的NAPI机制，但是产生的中断数量仍然很多。OVS DPDK以非常高的速率发送数据包到客户机。同时，QEMU的virtio队列的收发缓存数被限制在了默认的256与最大1024之间。结果，客户机必须以非常快的速度处理数据包。理想的实现就是使用DPDK的PMD驱动不停的轮询客户机端口进行数据包处理。
vhost user协议标准
参见QEMU代码库中文档：https://github.com/qemu/qemu/blob/master/docs/interop/vhost-user.txt
Vhost-user协议
===================
Copyright (c) 2014 Virtual Open Systems Sarl.
This work is licensed under the terms of the GNU GPL, version 2 or later.
See the COPYING file in the top-level directory.
===================
此协议旨在补充实现在Linux内核中的vhost的ioctl接口。实现了与同一宿主机中的用户进程交互建立virtqueue队列的控制平面。通过UNIX套接口消息中的附加数据字段来共享文件描述符。
协议定义了通信的两端：主和从。主时要共享其virtqueues队列的进程，即QEMU。从为virtqueues队列的消费者。
当前实现中QEMU作为主，从为运行在用户空间的软件交换机，如Snabbswitch。
主和从在通信时都可以作为客户端（主动连接）或者服务端（监听）。vhost user协议由两方组成：
- 主方 - QEMU
- 从方 - Open vSwitch或者其它软件交换机
vhost user各方都可运行在2中模式下：
- vhostuser-client - QEMU作为服务端，软件交换机作为客户端
- vhostuser - 软件交换机作为服务端，QEMU作为客户端。
vhost user实现基于内核的vhost架构，将所有特性实现在用户空间。
当QEMU客户机启动时，它将所有的客户机内存分配为共享的巨页内存。其操作系统的半虚拟化驱动virtio将保留这些巨页内存的一部分用作virtio环形缓存。这样OVS DPDK将可以直接读写客户机的virtio环形缓存。OVS DPDK和QEMU可通过此保留的内存空间交换网络数据包。
用户空间进程接收到客户机预先分配的共享内存文件描述符后，可直接存取与之关联的客户机内存空间中的vrings环结构。 (http://www.virtualopensystems.com/en/solutions/guides/snabbswitch-qemu/).
参见以下的VM虚拟机，模式为vhostuser：
$ /usr/libexec/qemu-kvm -name guest=instance-00000028,debug-threads=on -S -object secret,id=masterKey0,format=raw,file=/var/lib/libvirt/qemu/domain-58-instance-00000028/master-key.aes -machine pc-i440fx-rhel7.4.0,accel=kvm,usb=off,dump-guest-core=off -cpu Skylake-Client,ss=on,hypervisor=on,tsc_adjust=on,pdpe1gb=on,mpx=off,xsavec=off,xgetbv1=off -m 2048 -realtime mlock=off
-smp 8,sockets=4,cores=1,threads=2
-object memory-backend-file,id=ram-node0,prealloc=yes,mem-path=/dev/hugepages/libvirt/qemu/58-instance-00000028,share=yes,size=1073741824,host-nodes=0,policy=bind -numa node,nodeid=0,cpus=0-3,memdev=ram-node0
-object memory-backend-file,id=ram-node1,prealloc=yes,mem-path=/dev/hugepages/libvirt/qemu/58-instance-00000028,share=yes,size=1073741824,host-nodes=1,policy=bind
-numa node,nodeid=1,cpus=4-7,memdev=ram-node1 -uuid 48888226-7b6b-415c-bcf7-b278ba0bca62 -smbios type=1,manufacturer=Red Hat,product=OpenStack Compute,version=14.1.0-3.el7ost,serial=3d5e138a-8193-41e4-ac95-de9bfc1a3ef1,uuid=48888226-7b6b-415c-bcf7-b278ba0bca62,family=Virtual Machine -no-user-config -nodefaults -chardev socket,id=charmonitor,path=/var/lib/libvirt/qemu/domain-58-instance-00000028/monitor.sock,server,nowait -mon chardev=charmonitor,id=monitor,mode=control -rtc base=utc,driftfix=slew -global kvm-pit.lost_tick_policy=delay \
-no-hpet -no-shutdown -boot strict=on -device piix3-usb-uhci,id=usb,bus=pci.0,addr=0x1.0x2 -drive file=/var/lib/nova/instances/48888226-7b6b-415c-bcf7-b278ba0bca62/disk,format=qcow2,if=none,id=drive-virtio-disk0,cache=none
-device virtio-blk-pci,scsi=off,bus=pci.0,addr=0x4,drive=drive-virtio-disk0,id=virtio-disk0,bootindex=1 -chardev socket,id=charnet0,path=/var/run/openvswitch/vhuc26fd3c6-4b
-netdev vhost-user,chardev=charnet0,queues=8,id=hostnet0 \
-device virtio-net-pci,mq=on,vectors=18,netdev=hostnet0,id=net0,mac=fa:16:3e:52:30:73,bus=pci.0,addr=0x3 -add-fd set=0,fd=33 -chardev file,id=charserial0,path=/dev/fdset/0,append=on \
-device isa-serial,chardev=charserial0,id=serial0 -chardev pty,id=charserial1 \
-device isa-serial,chardev=charserial1,id=serial1 \
-device usb-tablet,id=input0,bus=usb.0,port=1 -vnc 172.16.2.10:1 -k en-us \
-device cirrus-vga,id=video0,bus=pci.0,addr=0x2 \
-device virtio-balloon-pci,id=balloon0,bus=pci.0,addr=0x5 -msg timestamp=on指定QEMU从巨页池中分配内存，并设置为共享内存。
-object memory-backend-file,id=ram-node0,prealloc=yes,mem-path=/dev/hugepages/libvirt/qemu/58-instance-00000028,share=yes,size=1073741824,host-nodes=0,policy=bind -numa node,nodeid=0,cpus=0-3,memdev=ram-node0 \
-object memory-backend-file,id=ram-node1,prealloc=yes,mem-path=/dev/hugepages/libvirt/qemu/58-instance-00000028,share=yes,size=1073741824,host-nodes=1,policy=bind尽管如此，简单的拷贝数据包到对方的缓存中还不足够。另外，vhost user协议使用一个UNIX套接口（vhu[a-f0-9-]）处理vswitch和QEMU之间的通信，包括在初始化过程中，和数据包拷贝到共享内存的virtio环中需要通知对方时。所以两者的交互包括基于控制通道（vhu）的创建操作和通知机制，与拷贝数据包的数据通道（直接内存访问）。
所述virtio机制要能工作，我们需要建立一个接口来初始化共享内存区域和交换event事件描述符。UNIX套接口提供的API接口可实现此要求。此套接口可用于初始化用户空间virtio传输（vhost-user），特别是：
* 初始化时确定Vrings，并且放入两个进程间的共享内存中；
* 使用eventfd映射到Vring事件。这样就可与QEMU/KVM中的实现相兼容，KVM可以关联客户机系统中virtio_pci驱动所触发事件与宿主机的eventfd（ioventfd和irqfd）文件描述符。
在两个进程间共享文件描述符与在一个进程和内核直接不相同。前者需要在UNIX套接口的sendmsg系统调用中设置SCM_RIGHTS标志。(http://www.virtualopensystems.com/en/solutions/guides/snabbswitch-qemu/)
vhostuser模式下，OVS创建vhu套接口，QEMU主动进行连接。vhostuser client模式下，QEMU创建vhu套接口，OVS进行连接。
在上面创建的vhostuser模式客户机实例中，指示QEMU连接一个类型为vhost-user的netdev到套接口/var/run/openvswitch/vhuc26fd3c6-4b:
-chardev socket,id=charnet0,path=/var/run/openvswitch/vhuc26fd3c6-4b \
-netdev vhost-user,chardev=charnet0,queues=8,id=hostnet0 \
-device virtio-net-pci,mq=on,vectors=18, \
netdev=hostnet0,id=net0,mac=fa:16:3e:52:30:73,bus=pci.0,addr=0x3使用lsof命令显示此套接口为OVS所创建：
[root@overcloud-0 ~]# lsof -nn | grep vhuc26fd3c6-4b | awk '{print $1}' | uniq当一方拷贝一个数据报文到共享内存的virtio环中时，另一方有两种选择：
- 类似(e.g. Linux kernel's NAPI)或者 (e.g. DPDK's PMD)的轮询队列，不需要通知就可取得新的数据报文；
- 非队列轮询，必须得到新报文到达的通知。
针对第二种情况，可通过独立的vhu套接口控制通道发送通知到客户机。通过交换eventfd文件描述符数据，控制通道可在用户空间实现中断。套接口的写操作要求系统调用，必将引起PMDs花费时间在内核空间。客户机可通过设置VRING_AVAIL_F_NO_INTERRUPT标志关闭控制通道中断通知。否则，当Open vSwitch网virtio环中填入新数据包时，将发送中断通知到客户机。
详情可参加此博客文章：http://blog.vmsplice.net/2011/09/qemu-internals-vhost-architecture.html
用户空间的vhost接口
vhost架构的一个惊人的特性是其并没有绑定在KVM上。其仅是一个用户空间接口并不依赖于KVM内核模块。这意味着其它的用户空间程序，
比如libpcap，如果要获得高性能I/O接口，理论上也可以使用vhost设备。
当客户机通知宿主机其已在virtqueue中填入了数据时，需要通知vhost的工作进程有数据要进行处理（对于内核的virtio-net驱动，
vhost工作进程为一个内核线程，名称为vhost-$pid，其中pid为QEMU的进程号）。既然vhost不依赖于KVM内核模块，二者就不能直接通信。
所以vhost实例创建了一个eventfd文件描述符，提供给vhost工作进程去监听。KVM内核模块的ioeventfd特性可将一个eventfd文件
描述符关联到一个特殊的客户机I/O操作上。QEMU用户空间在硬件寄存器VIRTIO_PCI_QUEUE_NOTIFY的I/O访问上注册了virtqueue的通知ioeventfd。
当客户机写VIRTIO_PCI_QUEUE_NOTIFY寄存器时将会发送virtqueue队列通知，vhost工作进程将接收到KVM内核模块通过ioeventfd发来的通知。
在vhost工作进程需要发送中断到客户机的反向路径上使用相同的方式。vhost通过写一个“call”文件描述符去通知客户机。
KVM内核模块的另一个特性irqfd中断描述符可使eventfd出发客户机中断。QEMU用户空间为virtio的PCI设备中断注册了一个irqfd文件描述符，
并将此irqfd交于vhost实例。vhost工作进程即可通过此“call”文件描述符去中断客户机。
最终，vhost实例仅了解到客户机的内存映射、kick通知eventfd文件描述符和call中断文件描述符。
更多细节，参考Linux内核中内核相关代码：
drivers/vhost/vhost.c - 通用vhost驱动代码
drivers/vhost/net.c - vhost-net网络设备驱动代码
virt/kvm/eventfd.c - ioeventfd事件和irqfd中断文件描述符实现
QEMU初始化vhost实例的用户空间代码：
hw/vhost.c - 通用vhost初始化代码
hw/vhost_net.c - vhost-net网络设备初始化代码数据通道-直接内存访问
virtqueue的内存映射
virtio官方标准定义了virtqueue的结构。
2.4 Virtqueues
virtio设备的大数据传输机制命名为virtqueue虚拟队列。每个设备可以有多个virtqueues，也可以没有
virtqueue队列。16位的队列大小参数指定了队列内成员的数量，也限定了队列的总大小。
每个virtqueue队列有三个部分组成：
Descriptor Table - 描述符表
Available Ring - 可用环
Used Ring - 已用环http://docs.oasis-open.org/virtio/virtio/v1.0/virtio-v1.0.html
virtio标志精确的定义了描述符表、可用环和已用环的结构。例如，可用环的定义：
2.4.6 virtqueue可用环结构
struct virtq_avail {
#define VIRTQ_AVAIL_F_NO_INTERRUPT 1
le16 flags;
le16 idx;
le16 ring[ /* Queue Size */ ];
le16 used_event; /* Only if VIRTIO_F_EVENT_IDX */
};
驱动程序使用可用环提供发送缓存给设备。其中每个环项指向一个描述符链的开头。可用环只能由驱动程序写，由设备读。
idx成员指示驱动程序将下一个描述符入口项放在了ring成员的哪个位置（不超过队列长度）。其从0开始增加。
传统的标准[Virtio PCI Draft]将此结构定义为vring_avail，将宏定义命名为
VRING_AVAIL_F_NO_INTERRUPT，但是本质结构都还是相同的。http://docs.oasis-open.org/virtio/virtio/v1.0/virtio-v1.0.html
DPDK的virtio标准实现代码，其也是使用传统virtio标准中的结构定义：
dpdk-18.08/drivers/net/virtio/virtio_ring.h
48 /* The Host uses this in used->flags to advise the Guest: don't kick me
49 * when you add a buffer. It's unreliable, so it's simply an
50 * optimization. Guest will still kick if it's out of buffers. */
51 #define VRING_USED_F_NO_NOTIFY 1
52 /* The Guest uses this in avail->flags to advise the Host: don't
53 * interrupt me when you consume a buffer. It's unreliable, so it's
54 * simply an optimization. */
55 #define VRING_AVAIL_F_NO_INTERRUPT 1
56
57 /* VirtIO ring descriptors: 16 bytes.
58 * These can chain together via "next". */
59 struct vring_desc {
60 uint64_t addr; /* Address (guest-physical). */
61 uint32_t len; /* Length. */
62 uint16_t flags; /* The flags as indicated above. */
63 uint16_t next; /* We chain unused descriptors via this. */
64 };
65
66 struct vring_avail {
67 uint16_t flags;
68 uint16_t idx;
69 uint16_t ring[0];
70 };
71
72 /* id is a 16bit index. uint32_t is used here for ids for padding reasons. */
73 struct vring_used_elem {
74 /* Index of start of used descriptor chain. */
75 uint32_t id;
76 /* Total length of the descriptor chain which was written to. */
77 uint32_t len;
78 };
79
80 struct vring_used {
81 uint16_t flags;
82 volatile uint16_t idx;
83 struct vring_used_elem ring[0];
84 };
85
86 struct vring {
87 unsigned int num;
88 struct vring_desc *desc;
89 struct vring_avail *avail;
90 struct vring_used *used;
91 };dpdk-18.08/lib/librte_vhost/vhost.h
90 /**
91 * Structure contains variables relevant to RX/TX virtqueues.
92 */
93 struct vhost_virtqueue {
94 union {
95 struct vring_desc *desc;
96 struct vring_packed_desc *desc_packed;
97 };
98 union {
99 struct vring_avail *avail;
100 struct vring_packed_desc_event *driver_event;
101 };
102 union {
103 struct vring_used *used;
104 struct vring_packed_desc_event *device_event;
105 };
106 uint32_t size;
107
108 uint16_t last_avail_idx;
109 uint16_t last_used_idx;
110 /* Last used index we notify to front end. */
111 uint16_t signalled_used;
112 bool signalled_used_valid;
113 #define VIRTIO_INVALID_EVENTFD (-1)
114 #define VIRTIO_UNINITIALIZED_EVENTFD (-2)
115
116 /* Backend value to determine if device should started/stopped */
117 int backend;
118 int enabled;
119 int access_ok;
120 rte_spinlock_t access_lock;
121
122 /* Used to notify the guest (trigger interrupt) */
123 int callfd;
124 /* Currently unused as polling mode is enabled */
125 int kickfd;
126
127 /* Physical address of used ring, for logging */
128 uint64_t log_guest_addr;
129
130 uint16_t nr_zmbuf;
131 uint16_t zmbuf_size;
132 uint16_t last_zmbuf_idx;
133 struct zcopy_mbuf *zmbufs;
134 struct zcopy_mbuf_list zmbuf_list;
135
136 union {
137 struct vring_used_elem *shadow_used_split;
138 struct vring_used_elem_packed *shadow_used_packed;
139 };
140 uint16_t shadow_used_idx;
141 struct vhost_vring_addr ring_addrs;
142
143 struct batch_copy_elem *batch_copy_elems;
144 uint16_t batch_copy_nb_elems;
145 bool used_wrap_counter;
146 bool avail_wrap_counter;
147
148 struct log_cache_entry log_cache[VHOST_LOG_CACHE_NR];
149 uint16_t log_cache_nb_elem;
150
151 rte_rwlock_t iotlb_lock;
152 rte_rwlock_t iotlb_pending_lock;
153 struct rte_mempool *iotlb_pool;
154 TAILQ_HEAD(, vhost_iotlb_entry) iotlb_list;
155 int iotlb_cache_nr;
156 TAILQ_HEAD(, vhost_iotlb_entry) iotlb_pending_list;
157 } __rte_cache_aligned;内存映射完成之后，DPDP就可像客户机的virtio-net驱动一样直接操作其共享内存中的同一结构了。
控制通道-UNIX套接口
QEMU与DPDK通过vhost user套接口交换消息。
DPDK与QEMU的通信遵照标准的vhost-user协议。
消息类型如下：
dpdk-18.08/lib/librte_vhost/vhost_user.h
27 typedef enum VhostUserRequest {
28 VHOST_USER_NONE = 0,
29 VHOST_USER_GET_FEATURES = 1,
30 VHOST_USER_SET_FEATURES = 2,
31 VHOST_USER_SET_OWNER = 3,
32 VHOST_USER_RESET_OWNER = 4,
33 VHOST_USER_SET_MEM_TABLE = 5,
34 VHOST_USER_SET_LOG_BASE = 6,
35 VHOST_USER_SET_LOG_FD = 7,
36 VHOST_USER_SET_VRING_NUM = 8,
37 VHOST_USER_SET_VRING_ADDR = 9,
38 VHOST_USER_SET_VRING_BASE = 10,
39 VHOST_USER_GET_VRING_BASE = 11,
40 VHOST_USER_SET_VRING_KICK = 12,
41 VHOST_USER_SET_VRING_CALL = 13,
42 VHOST_USER_SET_VRING_ERR = 14,
43 VHOST_USER_GET_PROTOCOL_FEATURES = 15,
44 VHOST_USER_SET_PROTOCOL_FEATURES = 16,
45 VHOST_USER_GET_QUEUE_NUM = 17,
46 VHOST_USER_SET_VRING_ENABLE = 18,
47 VHOST_USER_SEND_RARP = 19,
48 VHOST_USER_NET_SET_MTU = 20,
49 VHOST_USER_SET_SLAVE_REQ_FD = 21,
50 VHOST_USER_IOTLB_MSG = 22,
51 VHOST_USER_CRYPTO_CREATE_SESS = 26,
52 VHOST_USER_CRYPTO_CLOSE_SESS = 27,
53 VHOST_USER_MAX = 28
54 } VhostUserRequest;更详细的有关消息类型的信息参见QEMU源代码中的文件：
https://github.com/qemu/qemu/blob/master/docs/interop/vhost-user.txt
DPDK使用如下函数处理接收到的消息：
dpdk-18.08/lib/librte_vhost/vhost_user.c
1548 int
1549 vhost_user_msg_handler(int vid, int fd)
1550 {还有dpdk-18.08/lib/librte_vhost/vhost_user.c：
1406 /* return bytes# of read on success or negative val on failure. */
1407 static int
1408 read_vhost_message(int sockfd, struct VhostUserMsg *msg)
1409 {DPDK向外发送消息使用如下函数
dpdk-18.08/lib/librte_vhost/vhost_user.c
1436 static int
1437 send_vhost_message(int sockfd, struct VhostUserMsg *msg, int *fds, int fd_num)
1438 {
1439 if (!msg)
1440 return 0;
1441
1442 return send_fd_message(sockfd, (char *)msg,
1443 VHOST_USER_HDR_SIZE + msg->size, fds, fd_num);
1444 }QEMU与之相对应的接收函数为：
qemu-3.0.0/contrib/libvhost-user/libvhost-user.c
1218 static bool
1219 vu_process_message(VuDev *dev, VhostUserMsg *vmsg)QEMU对应的消息发送函数：
qemu-3.0.0/hw/virtio/vhost-user.c
297 /* most non-init callers ignore the error */
298 static int vhost_user_write(struct vhost_dev *dev, VhostUserMsg *msg,
299 int *fds, int fd_num)
300 {DPDK UNIX套接口的注册和消息交互
neutron控制Open vSwitch创建一个名称为vhuxxxxxxxx-xx的接口。在OVS内部，此名称保存在netdev结构体的成员name中（netdev->name）。
当创建vhost user接口时，Open vSwitch控制DPDK注册一个新的vhost-user UNIX套接口。套接口的路径为vhost_sock_dir加netdev->name加设备的dev->vhost_id。
通过设置RTE_VHOST_USER_CLIENT标志，OVS可请求创建vhost user套接口的客户端模式。
OVS函数netdev_dpdk_vhost_construct调用DPDK的rte_vhost_driver_register函数，其又调用vhost_user_create_server或者vhost_user_create_client函数创建套接口。默认使用前者创建服务端模式的套接口，如果设置了RTE_VHOST_USER_CLIENT标志，创建客户端模式套接口。
相关的函数调用关系如下：
OVS
netdev_dpdk_vhost_construct
(struct netdev *netdev)
|
|
DPDK V
rte_vhost_driver_register
(const char *path, uint64_t flags)
|
V
create_unix_socket
(struct vhost_user_socket *vsocket)
|
OVS
netdev_dpdk_vhost_construct
(struct netdev *netdev)
|
DPDK V
rte_vhost_driver_start
(const char *path))
-----------------------------------------------
| |
V |
vhost_user_start_server |
(struct vhost_user_socket *vsocket) |
| |
V V
vhost_user_server_new_connection vhost_user_start_client vhost_user_client_reconnect
(int fd, void *dat, int *remove __rte_unused) (struct vhost_user_socket *vsocket) (void *arg __rte_unused)
| | |
V V V
--------------------------------------------------------------------------------------------------
|
V
vhost_user_add_connection
(int fd, struct vhost_user_socket *vsocket)
|
V
vhost_user_read_cb
(int connfd, void *dat, int *remove)
|
V
vhost_user_msg_handlernetdev_dpdk_vhost_construct定义在文件openvswitch-2.9.2/lib/netdev-dpdk.c
1058 static int
1059 netdev_dpdk_vhost_construct(struct netdev *netdev)
1060 {
1061 struct netdev_dpdk *dev = netdev_dpdk_cast(netdev);
1062 const char *name = netdev->name;
1063 int err;
1064
1065 /* 'name' is appended to 'vhost_sock_dir' and used to create a socket in
1066 * the file system. '/' or '\' would traverse directories, so they're not
1067 * acceptable in 'name'. */
1068 if (strchr(name, '/') || strchr(name, '\\')) {
1069 VLOG_ERR("\"%s\" is not a valid name for a vhost-user port. "
1070 "A valid name must not include '/' or '\\'",
1071 name);
1072 return EINVAL;
1073 }
1074
1075 ovs_mutex_lock(&dpdk_mutex);
1076 /* Take the name of the vhost-user port and append it to the location where
1077 * the socket is to be created, then register the socket.
1078 */
1079 snprintf(dev->vhost_id, sizeof dev->vhost_id, "%s/%s",
1080 dpdk_get_vhost_sock_dir(), name);
1081
1082 dev->vhost_driver_flags &= ~RTE_VHOST_USER_CLIENT;
1083 err = rte_vhost_driver_register(dev->vhost_id, dev->vhost_driver_flags);
1084 if (err) {
1085 VLOG_ERR("vhost-user socket device setup failure for socket %s\n",
1086 dev->vhost_id);
1087 goto out;
1088 } else {
1089 fatal_signal_add_file_to_unlink(dev->vhost_id);
1090 VLOG_INFO("Socket %s created for vhost-user port %s\n",
1091 dev->vhost_id, name);
1092 }
1093
1094 err = rte_vhost_driver_callback_register(dev->vhost_id,
1095 &virtio_net_device_ops);
1096 if (err) {
1097 VLOG_ERR("rte_vhost_driver_callback_register failed for vhost user "
1098 "port: %s\n", name);
1099 goto out;
1100 }
1101
1102 err = rte_vhost_driver_disable_features(dev->vhost_id,
1103 1ULL << VIRTIO_NET_F_HOST_TSO4
1104 | 1ULL << VIRTIO_NET_F_HOST_TSO6
1105 | 1ULL << VIRTIO_NET_F_CSUM);
1106 if (err) {
1107 VLOG_ERR("rte_vhost_driver_disable_features failed for vhost user "
1108 "port: %s\n", name);
1109 goto out;
1110 }
1111
1112 err = rte_vhost_driver_start(dev->vhost_id);
1113 if (err) {
1114 VLOG_ERR("rte_vhost_driver_start failed for vhost user "
1115 "port: %s\n", name);
1116 goto out;
1117 }
1118
1119 err = vhost_common_construct(netdev);
1120 if (err) {
1121 VLOG_ERR("vhost_common_construct failed for vhost user "
1122 "port: %s\n", name);
1123 }
1124
1125 out:
1126 ovs_mutex_unlock(&dpdk_mutex);
1127 VLOG_WARN_ONCE("dpdkvhostuser ports are considered deprecated; "
1128 "please migrate to dpdkvhostuserclient ports.");
1129 return err;
1130 }netdev_dpdk_vhost_construct函数调用rte_vhost_driver_register。以下代码均定义在dpdk-18.08/lib/librte_vhost/socket.c
798 /*
799 * Register a new vhost-user socket; here we could act as server
800 * (the default case), or client (when RTE_VHOST_USER_CLIENT) flag
801 * is set.
802 */
803 int
804 rte_vhost_driver_register(const char *path, uint64_t flags)
805 {
867 if ((flags & RTE_VHOST_USER_CLIENT) != 0) {
868 vsocket->reconnect = !(flags & RTE_VHOST_USER_NO_RECONNECT);
869 if (vsocket->reconnect && reconn_tid == 0) {
870 if (vhost_user_reconnect_init() != 0)
871 goto out_mutex;
872 }
873 } else {
874 vsocket->is_server = true;
875 }
876 ret = create_unix_socket(vsocket);
877 if (ret < 0) {
878 goto out_mutex;
879 }netdev_dpdk_vhost_construct函数调用rte_vhost_driver_start。定义在dpdk-18.08/lib/librte_vhost/socket.c
1023 int
1024 rte_vhost_driver_start(const char *path)
1025 {
1059 if (vsocket->is_server)
1060 return vhost_user_start_server(vsocket);
1061 else
1062 return vhost_user_start_client(vsocket);
1063 }vhost_user_create_server调用vhost_user_server_new_connection:
以下的3个函数调用vhost_user_add_connection：
266 /* call back when there is new vhost-user connection from client */
267 static void
268 vhost_user_server_new_connection(int fd, void *dat, int *remove __rte_unused)
269 {
(...)
424 static void *
425 vhost_user_client_reconnect(void *arg __rte_unused)
426 {
(...)
494 static int
495 vhost_user_start_client(struct vhost_user_socket *vsocket)
496 {
(...)
194 static void
195 vhost_user_add_connection(int fd, struct vhost_user_socket *vsocket)
196 {vhost_user_add_connection接下来执行vhost_user_read_cb函数，其又调用vhost_user_msg_handler函数处理接收到的消息。
280 static void
281 vhost_user_read_cb(int connfd, void *dat, int *remove)
282 {
283 struct vhost_user_connection *conn = dat;
284 struct vhost_user_socket *vsocket = conn->vsocket;
285 int ret;
286
287 ret = vhost_user_msg_handler(conn->vid, connfd);
288 if (ret < 0) {
289 close(connfd);
290 *remove = 1;
291 vhost_destroy_device(conn->vid);
292
293 if (vsocket->notify_ops->destroy_connection)
294 vsocket->notify_ops->destroy_connection(conn->vid);
295
296 pthread_mutex_lock(&vsocket->conn_mutex);
297 TAILQ_REMOVE(&vsocket->conn_list, conn, next);
298 pthread_mutex_unlock(&vsocket->conn_mutex);
299
300 free(conn);
301
302 if (vsocket->reconnect) {
303 create_unix_socket(vsocket);
304 vhost_user_start_client(vsocket);
305 }
306 }
307 }dpdk-18.08/lib/librte_vhost/vhost_user.c
1548 int
1549 vhost_user_msg_handler(int vid, int fd)
1550 {
1551 struct virtio_net *dev;
1552 struct VhostUserMsg msg;
1553 struct rte_vdpa_device *vdpa_dev;
1554 int did = -1;
1555 int ret;
1556 int unlock_required = 0;
1557 uint32_t skip_master = 0;
1558
1559 dev = get_device(vid);
1560 if (dev == NULL)
1561 return -1;
1562
1563 if (!dev->notify_ops) {
1564 dev->notify_ops = vhost_driver_callback_get(dev->ifname);
1565 if (!dev->notify_ops) {
1566 RTE_LOG(ERR, VHOST_CONFIG,
1567 "failed to get callback ops for driver %s\n",
1568 dev->ifname);
1569 return -1;
1570 }
1571 }
1572
1573 ret = read_vhost_message(fd, &msg);
1574 if (ret <= 0 || msg.request.master >= VHOST_USER_MAX) {
1575 if (ret < 0)
1576 RTE_LOG(ERR, VHOST_CONFIG,
1577 "vhost read message failed\n");
1578 else if (ret == 0)
1579 RTE_LOG(INFO, VHOST_CONFIG,
1580 "vhost peer closed\n");
1581 else
1582 RTE_LOG(ERR, VHOST_CONFIG,
1583 "vhost read incorrect message\n");
1584
1585 return -1;
1586 }
1587
1588 ret = 0;
1589 if (msg.request.master != VHOST_USER_IOTLB_MSG)
1590 RTE_LOG(INFO, VHOST_CONFIG, "read message %s\n",
1591 vhost_message_str[msg.request.master]);
1592 else
1593 RTE_LOG(DEBUG, VHOST_CONFIG, "read message %s\n",
1594 vhost_message_str[msg.request.master]);
1595
1596 ret = vhost_user_check_and_alloc_queue_pair(dev, &msg);
1597 if (ret < 0) {
1598 RTE_LOG(ERR, VHOST_CONFIG,
1599 "failed to alloc queue\n");
1600 return -1;
1601 }
1602
1603 /*
1604 * Note: we don't lock all queues on VHOST_USER_GET_VRING_BASE
1605 * and VHOST_USER_RESET_OWNER, since it is sent when virtio stops
1606 * and device is destroyed. destroy_device waits for queues to be
1607 * inactive, so it is safe. Otherwise taking the access_lock
1608 * would cause a dead lock.
1609 */
1610 switch (msg.request.master) {
1611 case VHOST_USER_SET_FEATURES:
1612 case VHOST_USER_SET_PROTOCOL_FEATURES:
1613 case VHOST_USER_SET_OWNER:
1614 case VHOST_USER_SET_MEM_TABLE:
1615 case VHOST_USER_SET_LOG_BASE:
1616 case VHOST_USER_SET_LOG_FD:
1617 case VHOST_USER_SET_VRING_NUM:
1618 case VHOST_USER_SET_VRING_ADDR:
1619 case VHOST_USER_SET_VRING_BASE:
1620 case VHOST_USER_SET_VRING_KICK:
1621 case VHOST_USER_SET_VRING_CALL:
1622 case VHOST_USER_SET_VRING_ERR:
1623 case VHOST_USER_SET_VRING_ENABLE:
1624 case VHOST_USER_SEND_RARP:
1625 case VHOST_USER_NET_SET_MTU:
1626 case VHOST_USER_SET_SLAVE_REQ_FD:
1627 vhost_user_lock_all_queue_pairs(dev);
1628 unlock_required = 1;
1629 break;
1630 default:
1631 break;
1632
1633 }
1634
1635 if (dev->extern_ops.pre_msg_handle) {
1636 uint32_t need_reply;
1637
1638 ret = (*dev->extern_ops.pre_msg_handle)(dev->vid,
1639 (void *)&msg, &need_reply, &skip_master);
1640 if (ret < 0)
1641 goto skip_to_reply;
1642
1643 if (need_reply)
1644 send_vhost_reply(fd, &msg);
1645
1646 if (skip_master)
1647 goto skip_to_post_handle;
1648 }
1649
1650 switch (msg.request.master) {
1651 case VHOST_USER_GET_FEATURES:
1652 msg.payload.u64 = vhost_user_get_features(dev);
1653 msg.size = sizeof(msg.payload.u64);
1654 send_vhost_reply(fd, &msg);
1655 break;
1656 case VHOST_USER_SET_FEATURES:
1657 ret = vhost_user_set_features(dev, msg.payload.u64);
1658 if (ret)
1659 return -1;
1660 break;
1661
1662 case VHOST_USER_GET_PROTOCOL_FEATURES:
1663 vhost_user_get_protocol_features(dev, &msg);
1664 send_vhost_reply(fd, &msg);
1665 break;
1666 case VHOST_USER_SET_PROTOCOL_FEATURES:
1667 vhost_user_set_protocol_features(dev, msg.payload.u64);
1668 break;
1669
1670 case VHOST_USER_SET_OWNER:
1671 vhost_user_set_owner();
1672 break;
1673 case VHOST_USER_RESET_OWNER:
1674 vhost_user_reset_owner(dev);
1675 break;
1676
1677 case VHOST_USER_SET_MEM_TABLE:
1678 ret = vhost_user_set_mem_table(&dev, &msg);
1679 break;
1680
1681 case VHOST_USER_SET_LOG_BASE:
1682 vhost_user_set_log_base(dev, &msg);
1683
1684 /* it needs a reply */
1685 msg.size = sizeof(msg.payload.u64);
1686 send_vhost_reply(fd, &msg);
1687 break;
1688 case VHOST_USER_SET_LOG_FD:
1689 close(msg.fds[0]);
1690 RTE_LOG(INFO, VHOST_CONFIG, "not implemented.\n");
1691 break;
1692
1693 case VHOST_USER_SET_VRING_NUM:
1694 vhost_user_set_vring_num(dev, &msg);
1695 break;
1696 case VHOST_USER_SET_VRING_ADDR:
1697 vhost_user_set_vring_addr(&dev, &msg);
1698 break;
1699 case VHOST_USER_SET_VRING_BASE:
1700 vhost_user_set_vring_base(dev, &msg);
1701 break;
1702
1703 case VHOST_USER_GET_VRING_BASE:
1704 vhost_user_get_vring_base(dev, &msg);
1705 msg.size = sizeof(msg.payload.state);
1706 send_vhost_reply(fd, &msg);
1707 break;
1708
1709 case VHOST_USER_SET_VRING_KICK:
1710 vhost_user_set_vring_kick(&dev, &msg);
1711 break;
1712 case VHOST_USER_SET_VRING_CALL:
1713 vhost_user_set_vring_call(dev, &msg);
1714 break;
1715
1716 case VHOST_USER_SET_VRING_ERR:
1717 if (!(msg.payload.u64 & VHOST_USER_VRING_NOFD_MASK))
1718 close(msg.fds[0]);
1719 RTE_LOG(INFO, VHOST_CONFIG, "not implemented\n");
1720 break;
1721
1722 case VHOST_USER_GET_QUEUE_NUM:
1723 msg.payload.u64 = (uint64_t)vhost_user_get_queue_num(dev);
1724 msg.size = sizeof(msg.payload.u64);
1725 send_vhost_reply(fd, &msg);
1726 break;
1727
1728 case VHOST_USER_SET_VRING_ENABLE:
1729 vhost_user_set_vring_enable(dev, &msg);
1730 break;
1731 case VHOST_USER_SEND_RARP:
1732 vhost_user_send_rarp(dev, &msg);
1733 break;
1734
1735 case VHOST_USER_NET_SET_MTU:
1736 ret = vhost_user_net_set_mtu(dev, &msg);
1737 break;
1738
1739 case VHOST_USER_SET_SLAVE_REQ_FD:
1740 ret = vhost_user_set_req_fd(dev, &msg);
1741 break;
1742
1743 case VHOST_USER_IOTLB_MSG:
1744 ret = vhost_user_iotlb_msg(&dev, &msg);
1745 break;
1746
1747 default:
1748 ret = -1;
1749 break;
1750 }
1751
1752 skip_to_post_handle:
1753 if (dev->extern_ops.post_msg_handle) {
1754 uint32_t need_reply;
1755
1756 ret = (*dev->extern_ops.post_msg_handle)(
1757 dev->vid, (void *)&msg, &need_reply);
1758 if (ret < 0)
1759 goto skip_to_reply;
1760
1761 if (need_reply)
1762 send_vhost_reply(fd, &msg);
1763 }
1764
1765 skip_to_reply:
1766 if (unlock_required)
1767 vhost_user_unlock_all_queue_pairs(dev);
1768
1769 if (msg.flags & VHOST_USER_NEED_REPLY) {
1770 msg.payload.u64 = !!ret;
1771 msg.size = sizeof(msg.payload.u64);
1772 send_vhost_reply(fd, &msg);
1773 }
1774
1775 if (!(dev->flags & VIRTIO_DEV_RUNNING) && virtio_is_ready(dev)) {
1776 dev->flags |= VIRTIO_DEV_READY;
1777
1778 if (!(dev->flags & VIRTIO_DEV_RUNNING)) {
1779 if (dev->dequeue_zero_copy) {
1780 RTE_LOG(INFO, VHOST_CONFIG,
1781 "dequeue zero copy is enabled\n");
1782 }
1783
1784 if (dev->notify_ops->new_device(dev->vid) == 0)
1785 dev->flags |= VIRTIO_DEV_RUNNING;
1786 }
1787 }
1788
1789 did = dev->vdpa_dev_id;
1790 vdpa_dev = rte_vdpa_get_device(did);
1791 if (vdpa_dev && virtio_is_ready(dev) &&
1792 !(dev->flags & VIRTIO_DEV_VDPA_CONFIGURED) &&
1793 msg.request.master == VHOST_USER_SET_VRING_ENABLE) {
1794 if (vdpa_dev->ops->dev_conf)
1795 vdpa_dev->ops->dev_conf(dev->vid);
1796 dev->flags |= VIRTIO_DEV_VDPA_CONFIGURED;
1797 if (vhost_user_host_notifier_ctrl(dev->vid, true) != 0) {
1798 RTE_LOG(INFO, VHOST_CONFIG,
1799 "(%d) software relay is used for vDPA, performance may be low.\n",
1800 dev->vid);
1801 }
1802 }
1803
1804 return 0;
1805 }virtio告知DPDK共享内存的virtio queues内存地址
DPDK使用函数vhost_user_set_vring_addr将virtio的描述符、已用环和可用环地址转化为DPDK自身的地址空间。
dpdk-18.08/lib/librte_vhost/vhost_user.c
607 /*
608 * The virtio device sends us the desc, used and avail ring addresses.
609 * This function then converts these to our address space.
610 */
611 static int
612 vhost_user_set_vring_addr(struct virtio_net **pdev, VhostUserMsg *msg)
613 {
614 struct vhost_virtqueue *vq;
615 struct vhost_vring_addr *addr = &msg->payload.addr;
616 struct virtio_net *dev = *pdev;
617
618 if (dev->mem == NULL)
619 return -1;
620
621 /* addr->index refers to the queue index. The txq 1, rxq is 0. */
622 vq = dev->virtqueue[msg->payload.addr.index];
623
624 /*
625 * Rings addresses should not be interpreted as long as the ring is not
626 * started and enabled
627 */
628 memcpy(&vq->ring_addrs, addr, sizeof(*addr));
629
630 vring_invalidate(dev, vq);
631
632 if (vq->enabled && (dev->features &
633 (1ULL << VHOST_USER_F_PROTOCOL_FEATURES))) {
634 dev = translate_ring_addresses(dev, msg->payload.addr.index);
635 if (!dev)
636 return -1;
637
638 *pdev = dev;
639 }
640
641 return 0;
642 }只有在通过控制通道vhu套接口接收到VHOST_USER_SET_VRING_ADDR类型消息时，设置内存地址。
dpdk-18.08/lib/librte_vhost/vhost_user.c
1548 int
1549 vhost_user_msg_handler(int vid, int fd)
1550 {
(...)
1650 switch (msg.request.master) {
(...)
1696 case VHOST_USER_SET_VRING_ADDR:
1697 vhost_user_set_vring_addr(&dev, &msg);
1698 break;实际上，QEMU中有一个与DPDK的消息处理函数类型的处理函数。
qemu-3.0.0/contrib/libvhost-user/libvhost-user.c
1218 static bool
1219 vu_process_message(VuDev *dev, VhostUserMsg *vmsg)
1220 {
(...)
1244 switch (vmsg->request) {
(...)
1265 case VHOST_USER_SET_VRING_ADDR:
1266 return vu_set_vring_addr_exec(dev, vmsg);
(...) 显然，QEMU中需要有函数通过UNIX套接口发送内存地址信息到DPDK中。
qemu-3.0.0/hw/virtio/vhost-user.c
588 static int vhost_user_set_vring_addr(struct vhost_dev *dev,
589 struct vhost_vring_addr *addr)
590 {
591 VhostUserMsg msg = {
592 .hdr.request = VHOST_USER_SET_VRING_ADDR,
593 .hdr.flags = VHOST_USER_VERSION,
594 .payload.addr = *addr,
595 .hdr.size = sizeof(msg.payload.addr),
596 };
597
598 if (vhost_user_write(dev, &msg, NULL, 0) < 0) {
599 return -1;
600 }
601
602 return 0;
603 }OVS DPDK发送数据包到客户机与发送丢包
OVS DPDK中向客户机发送数据包的函数为__netdev_dpdk_vhost_send，位于文件openvswitch-2.9.2/lib/netdev-dpdk.c。
OVS发送程序，在空间用完后，仍会尝试发送VHOST_ENQ_RETRY_NUM (默认8)次。如果在第一次尝试发送中，没有任何数据包发送成功（无数据包写入共享内存的环中），或者超过了VHOST_ENQ_RETRY_NUM宏限定的次数，剩余的数据包将被丢弃（批量发送最大可由32个数据包组成）。
2072 do {
2073 int vhost_qid = qid * VIRTIO_QNUM + VIRTIO_RXQ;
2074 unsigned int tx_pkts;
2075
2076 tx_pkts = rte_vhost_enqueue_burst(vid, vhost_qid, cur_pkts, cnt);
2077 if (OVS_LIKELY(tx_pkts)) {
2078 /* Packets have been sent.*/
2079 cnt -= tx_pkts;
2080 /* Prepare for possible retry.*/
2081 cur_pkts = &cur_pkts[tx_pkts];
2082 } else {
2083 /* No packets sent - do not retry.*/
2084 break;
2085 }
2086 } while (cnt && (retries++ <= VHOST_ENQ_RETRY_NUM));
2087
（...）
2094
2095 out:
2096 for (i = 0; i < total_pkts - dropped; i++) {
2097 dp_packet_delete(pkts[i]);
2098 }客户机接收中断处理
当OVS DPDK将新的数据包填入virtio环中时，有以下两种情形：
- 客户机没有在轮询其队列，需要告知其新数据包的到达；
- 客户机正在轮询队列，不需要告知新数据包的到达。
如果客户机使用Linux内核网络协议栈，内核中负责接收报文的NAPI机制混合使用中断和轮询模式。客户机OS开始工作在中断模式，一直到第一个中断进来。此时，CPU快速响应中断，调度内核软中断ksoftirqd线程处理，同时禁止后续中断。
ksoftirqd运行时，尝试处理尽可能多的数据包，但是不能超出netdev_budget限定的数量。如果队列中还有更多的数据包，ksoftirqd线程将重新调度自身，继续处理数据包，直到没有可用的数据包为止。此过程中一直是轮询处理，中断处于关闭状态。处理完数据包之后，ksoftirqd线程停止轮询，重新打开中断，等待下一个数据包到来的中断发生。
当客户机轮询时，CPU的caches高速缓存利用率非常高，避免了额外的延时。宿主机和客户机中合适的进程在运行，进一步降低了延时。另外的，宿主机发送中断IRQ到客户机时，需要对UNIX套接口写操作（系统调用），非常耗时，增加了额外的延时和开销。
作为NFV应用的一部分，客户机中运行DPDK的优势在于其PMD驱动处理流量的方式。PMD驱动工作在轮询模式，关闭了系统中断，OVS DPDK不再需要给客户机发送中断通知。OVS DPDK节省了写UNIX套接口的操作，不在需要执行内核系统调用。OVS DPDK得以一直运行在用户空间，客户机也可以省去处理由控制通道而来的中断，快速运行。
如果没有设置VRING_AVAIL_F_NO_INTERRUPT标志，表明客户机可以接收中断。到客户机的中断通过callfd和操作系统的eventfd组件实现。
客户机的OS可以启用或禁用中断。当客户机禁用virtio接口的中断时，virtio-net驱动通过宏VRING_AVAIL_F_NO_INTERRUPT实现。此宏在DPDK和QEMU中都有定义：
[root@overcloud-0 SOURCES]# grep VRING_AVAIL_F_NO_INTERRUPT -R | grep def
dpdk-18.08/drivers/net/virtio/virtio_ring.h:#define VRING_AVAIL_F_NO_INTERRUPT 1
dpdk-18.08/drivers/crypto/virtio/virtio_ring.h:#define VRING_AVAIL_F_NO_INTERRUPT 1[root@overcloud-0 qemu]# grep AVAIL_F_NO_INTERRUPT -R -i | grep def
qemu-3.0.0/include/standard-headers/linux/virtio_ring.h:#define VRING_AVAIL_F_NO_INTERRUPT 1
qemu-3.0.0/roms/seabios/src/hw/virtio-ring.h:#define VRING_AVAIL_F_NO_INTERRUPT 1
qemu-3.0.0/roms/ipxe/src/include/ipxe/virtio-ring.h:#define VRING_AVAIL_F_NO_INTERRUPT 1
qemu-3.0.0/roms/seabios-hppa/src/hw/virtio-ring.h:#define VRING_AVAIL_F_NO_INTERRUPT 1
qemu-3.0.0/roms/SLOF/lib/libvirtio/virtio.h:#define VRING_AVAIL_F_NO_INTERRUPT 1一旦vq->avail->flags中的VRING_AVAIL_F_NO_INTERRUPT标志位设置，指示DPDK不要发送中断到客户机。dpdk-18.08/lib/librte_vhost/vhost.h
666 static __rte_always_inline void
667 vhost_vring_call_split(struct virtio_net *dev, struct vhost_virtqueue *vq)
668 {
669 /* Flush used->idx update before we read avail->flags. */
670 rte_smp_mb();
671
672 /* Don't kick guest if we don't reach index specified by guest. */
673 if (dev->features & (1ULL << VIRTIO_RING_F_EVENT_IDX)) {
674 uint16_t old = vq->signalled_used;
675 uint16_t new = vq->last_used_idx;
676
677 VHOST_LOG_DEBUG(VHOST_DATA, "%s: used_event_idx=%d, old=%d, new=%d\n",
678 __func__,
679 vhost_used_event(vq),
680 old, new);
681 if (vhost_need_event(vhost_used_event(vq), new, old)
682 && (vq->callfd >= 0)) {
683 vq->signalled_used = vq->last_used_idx;
684 eventfd_write(vq->callfd, (eventfd_t) 1);
685 }
686 } else {
687 /* Kick the guest if necessary. */
688 if (!(vq->avail->flags & VRING_AVAIL_F_NO_INTERRUPT)
689 && (vq->callfd >= 0))
690 eventfd_write(vq->callfd, (eventfd_t)1);
691 }
692 }如前所述，PMD驱动不需要执行写UNIX套接口的系统调用了。
OVS DPDK发送数据包到客户机-代码详情
2044 static void
2045 __netdev_dpdk_vhost_send(struct netdev *netdev, int qid,
2046 struct dp_packet **pkts, int cnt)
2047 {
2048 struct netdev_dpdk *dev = netdev_dpdk_cast(netdev);
2049 struct rte_mbuf **cur_pkts = (struct rte_mbuf **) pkts;
2050 unsigned int total_pkts = cnt;
2051 unsigned int dropped = 0;
2052 int i, retries = 0;
2053 int vid = netdev_dpdk_get_vid(dev);
2054
2055 qid = dev->tx_q[qid % netdev->n_txq].map;
2056
2057 if (OVS_UNLIKELY(vid < 0 || !dev->vhost_reconfigured || qid < 0
2058 || !(dev->flags & NETDEV_UP))) {
2059 rte_spinlock_lock(&dev->stats_lock);
2060 dev->stats.tx_dropped+= cnt;
2061 rte_spinlock_unlock(&dev->stats_lock);
2062 goto out;
2063 }
2064
2065 rte_spinlock_lock(&dev->tx_q[qid].tx_lock);
2066
2067 cnt = netdev_dpdk_filter_packet_len(dev, cur_pkts, cnt);
2068 /* Check has QoS has been configured for the netdev */
2069 cnt = netdev_dpdk_qos_run(dev, cur_pkts, cnt, true);
2070 dropped = total_pkts - cnt;
2071
2072 do {
2073 int vhost_qid = qid * VIRTIO_QNUM + VIRTIO_RXQ;
2074 unsigned int tx_pkts;
2075
2076 tx_pkts = rte_vhost_enqueue_burst(vid, vhost_qid, cur_pkts, cnt);
2077 if (OVS_LIKELY(tx_pkts)) {
2078 /* Packets have been sent.*/
2079 cnt -= tx_pkts;
2080 /* Prepare for possible retry.*/
2081 cur_pkts = &cur_pkts[tx_pkts];
2082 } else {
2083 /* No packets sent - do not retry.*/
2084 break;
2085 }
2086 } while (cnt && (retries++ <= VHOST_ENQ_RETRY_NUM));
2087
2088 rte_spinlock_unlock(&dev->tx_q[qid].tx_lock);
2089
2090 rte_spinlock_lock(&dev->stats_lock);
2091 netdev_dpdk_vhost_update_tx_counters(&dev->stats, pkts, total_pkts,
2092 cnt + dropped);
2093 rte_spinlock_unlock(&dev->stats_lock);
2094
2095 out:
2096 for (i = 0; i < total_pkts - dropped; i++) {
2097 dp_packet_delete(pkts[i]);
2098 }
2099 }rte_vhost_enqueue_burst函数来自于DPDK的vhost库。
[root@overcloud-0 src]# grep rte_vhost_enqueue_burst dpdk-16.08/ -R
dpdk-18.08/examples/vhost/main.c: ret = rte_vhost_enqueue_burst(dst_vdev->vid, VIRTIO_RXQ, &m, 1);
dpdk-18.08/examples/vhost/main.c: enqueue_count = rte_vhost_enqueue_burst(vdev->vid, VIRTIO_RXQ,
dpdk-18.08/examples/tep_termination/vxlan_setup.c: ret = rte_vhost_enqueue_burst(vid, VIRTIO_RXQ, pkts_valid, count);
dpdk-18.08/tags:rte_vhost_enqueue_burst lib/librte_vhost/virtio_net.c /^rte_vhost_enqueue_burst(int vid, uint16_t queue_id,$/;" f
dpdk-18.08/lib/librte_vhost/rte_vhost_version.map: rte_vhost_enqueue_burst;
dpdk-18.08/lib/librte_vhost/rte_vhost.h:uint16_t rte_vhost_enqueue_burst(int vid, uint16_t queue_id,
dpdk-18.08/lib/librte_vhost/virtio_net.c:rte_vhost_enqueue_burst(int vid, uint16_t queue_id,
dpdk-18.08/drivers/net/vhost/rte_eth_vhost.c: nb_pkts = rte_vhost_enqueue_burst(r->vid, r->virtqueue_id,
dpdk-18.08/doc/guides/prog_guide/vhost_lib.rst:* ``rte_vhost_enqueue_burst(vid, queue_id, pkts, count)``
dpdk-18.08/doc/guides/rel_notes/release_16_07.rst:* The function ``rte_vhost_enqueue_burst`` no longer supports concurrent enqueuingdpdk-18.08/lib/librte_vhost/rte_vhost.h
492 /**
493 * This function adds buffers to the virtio devices RX virtqueue. Buffers can
494 * be received from the physical port or from another virtual device. A packet
495 * count is returned to indicate the number of packets that were successfully
496 * added to the RX queue.
497 * @param vid
498 * vhost device ID
499 * @param queue_id
500 * virtio queue index in mq case
501 * @param pkts
502 * array to contain packets to be enqueued
503 * @param count
504 * packets num to be enqueued
505 * @return
506 * num of packets enqueued
507 */
508 uint16_t rte_vhost_enqueue_burst(int vid, uint16_t queue_id,
509 struct rte_mbuf **pkts, uint16_t count);dpdk-18.08/lib/librte_vhost/virtio_net.c
932 uint16_t
933 rte_vhost_enqueue_burst(int vid, uint16_t queue_id,
934 struct rte_mbuf **pkts, uint16_t count)
935 {
936 struct virtio_net *dev = get_device(vid);
937
938 if (!dev)
939 return 0;
940
941 if (unlikely(!(dev->flags & VIRTIO_DEV_BUILTIN_VIRTIO_NET))) {
942 RTE_LOG(ERR, VHOST_DATA,
943 "(%d) %s: built-in vhost net backend is disabled.\n",
944 dev->vid, __func__);
945 return 0;
946 }
947
948 return virtio_dev_rx(dev, queue_id, pkts, count);
949 }virtio_dev_rx_packed函数和virtio_dev_rx_split函数都将数据包发送到客户机，并根据设置决定是否发送中断通知（write系统调用）。
dpdk-18.08/lib/librte_vhost/virtio_net.c
886 static __rte_always_inline uint32_t
887 virtio_dev_rx(struct virtio_net *dev, uint16_t queue_id,
888 struct rte_mbuf **pkts, uint32_t count)
889 {
890 struct vhost_virtqueue *vq;
(...)
917 if (vq_is_packed(dev))
918 count = virtio_dev_rx_packed(dev, vq, pkts, count);
919 else
920 count = virtio_dev_rx_split(dev, vq, pkts, count);
(...)在virtio_dev_rx函数中：
913 count = RTE_MIN((uint32_t)MAX_PKT_BURST, count);
914 if (count == 0)
915 goto out;发送数据包数量设置为MAX_PKT_BURST宏与空闲项数量（count）两者中的较小值。
最后，根据发送的数据包数量增加已用索引的值。
833 static __rte_always_inline uint32_t
834 virtio_dev_rx_packed(struct virtio_net *dev, struct vhost_virtqueue *vq,
835 struct rte_mbuf **pkts, uint32_t count)
836 {
(...)
841 for (pkt_idx = 0; pkt_idx < count; pkt_idx++) {
(...)
862 if (copy_mbuf_to_desc(dev, vq, pkts[pkt_idx],
863 buf_vec, nr_vec,
864 num_buffers) < 0) {
865 vq->shadow_used_idx -= num_buffers;
866 break;
867 }
868
869 vq->last_avail_idx += nr_descs;
870 if (vq->last_avail_idx >= vq->size) {
871 vq->last_avail_idx -= vq->size;
872 vq->avail_wrap_counter ^= 1;
873 }
781 static __rte_always_inline uint32_t
782 virtio_dev_rx_split(struct virtio_net *dev, struct vhost_virtqueue *vq,
783 struct rte_mbuf **pkts, uint32_t count)
784 {
793 for (pkt_idx = 0; pkt_idx < count; pkt_idx++) {
794 uint32_t pkt_len = pkts[pkt_idx]->pkt_len + dev->vhost_hlen;
795 uint16_t nr_vec = 0;
(...)
813 if (copy_mbuf_to_desc(dev, vq, pkts[pkt_idx],
814 buf_vec, nr_vec,
815 num_buffers) < 0) {
816 vq->shadow_used_idx -= num_buffers;
817 break;
818 }
819
820 vq->last_avail_idx += num_buffers;
821 }数据包通过函数copy_mbuf_to_desc拷贝到客户机的内存中。最后，根据配置决定是否发送中断通知，参见函数vhost_vring_call_split和vhost_vring_call_packed。
666 static __rte_always_inline void
667 vhost_vring_call_split(struct virtio_net *dev, struct vhost_virtqueue *vq)
668 {
(...)
672 /* Don't kick guest if we don't reach index specified by guest. */
673 if (dev->features & (1ULL << VIRTIO_RING_F_EVENT_IDX)) {
674 uint16_t old = vq->signalled_used;
675 uint16_t new = vq->last_used_idx;
(...)
681 if (vhost_need_event(vhost_used_event(vq), new, old)
682 && (vq->callfd >= 0)) {
683 vq->signalled_used = vq->last_used_idx;
684 eventfd_write(vq->callfd, (eventfd_t) 1);
685 }
686 } else {
687 /* Kick the guest if necessary. */
688 if (!(vq->avail->flags & VRING_AVAIL_F_NO_INTERRUPT)
689 && (vq->callfd >= 0))
690 eventfd_write(vq->callfd, (eventfd_t)1);
691 }
692 }
694 static __rte_always_inline void
695 vhost_vring_call_packed(struct virtio_net *dev, struct vhost_virtqueue *vq)
696 {
(...)
703 if (!(dev->features & (1ULL << VIRTIO_RING_F_EVENT_IDX))) {
704 if (vq->driver_event->flags !=
705 VRING_EVENT_F_DISABLE)
706 kick = true;
707 goto kick;
708 }
(...)
740 kick:
741 if (kick)
742 eventfd_write(vq->callfd, (eventfd_t)1);
743 }
END
本文章为转载内容，我们尊重原作者对文章享有的著作权。如有内容错误或侵权问题，欢迎原作者联系我们进行内容更正或删除文章。
