ovs vswitchd的启动

vswitchd启动代码可参考ovs-vswitchd.c的main函数，其中最重要的两个函数是bridge_run以及netdev_run

bridge_run

void bridge_run() {.../* Initialize the ofproto library.  This only needs to run once, but* it must be done after the configuration is set.  If the* initialization has already occurred, bridge_init_ofproto()* returns immediately. */bridge_init_ofproto(cfg);bridge_run__();...
}

bridge_init_ofproto用于初始化ofproto-dpif组件，首先会从ovsdb记录中汇总bridge, port, interfaces等信息，之后针对这些调用ofproto_init

/* Must be called to initialize the ofproto library.** The caller may pass in 'iface_hints', which contains an shash of* "iface_hint" elements indexed by the interface's name.  The provider* may use these hints to describe the startup configuration in order to* reinitialize its state.  The caller owns the provided data, so a* provider will make copies of anything required.  An ofproto provider* will remove any existing state that is not described by the hint, and* may choose to remove it all. */
void
ofproto_init(const struct shash *iface_hints)
{struct shash_node *node;size_t i;ofproto_class_register(&ofproto_dpif_class);/* Make a local copy, since we don't own 'iface_hints' elements. */SHASH_FOR_EACH(node, iface_hints) {const struct iface_hint *orig_hint = node->data;struct iface_hint *new_hint = xmalloc(sizeof *new_hint);const char *br_type = ofproto_normalize_type(orig_hint->br_type);new_hint->br_name = xstrdup(orig_hint->br_name);new_hint->br_type = xstrdup(br_type);new_hint->ofp_port = orig_hint->ofp_port;shash_add(&init_ofp_ports, node->name, new_hint);}for (i = 0; i < n_ofproto_classes; i++) {ofproto_classes[i]->init(&init_ofp_ports);}ofproto_unixctl_init();
}

ofproto_init首先注册ofproto_dpif_class，这个是ovs的ofproto-dpif的默认实现，之后对已经注册的所有ofproto_classes，都会调用ofproto_class->init函数。
ofproto_dpif_class是ovs的ofproto实现，在ofproto/ofproto-dpif.c中，这里先介绍init函数

static void
init(const struct shash *iface_hints)
{struct shash_node *node;/* Make a local copy, since we don't own 'iface_hints' elements. */SHASH_FOR_EACH(node, iface_hints) {const struct iface_hint *orig_hint = node->data;struct iface_hint *new_hint = xmalloc(sizeof *new_hint);new_hint->br_name = xstrdup(orig_hint->br_name);new_hint->br_type = xstrdup(orig_hint->br_type);new_hint->ofp_port = orig_hint->ofp_port;shash_add(&init_ofp_ports, node->name, new_hint);}ofproto_unixctl_init();udpif_init();
}

ofproto_unixctl_init，udpif_init都通过unixctl_command_register注册了多个ovs-appctl的命令参数，

bridge_run___最终都是调用的ofproto_dpif_class里注册的函数

static void
bridge_run__(void)
{struct bridge *br;struct sset types;const char *type;/* Let each datapath type do the work that it needs to do. */sset_init(&types);ofproto_enumerate_types(&types);SSET_FOR_EACH (type, &types) {ofproto_type_run(type);}sset_destroy(&types);/* Let each bridge do the work that it needs to do. */HMAP_FOR_EACH (br, node, &all_bridges) {ofproto_run(br->ofproto);}
}

ofproto_enumerate_types实际调用的是ofproto_dpif_class里注册的enumerate_types函数，里面实际调用了dp_enumerate_types

dp_enumerate_types(struct sset *types)
{struct shash_node *node;dp_initialize();ovs_mutex_lock(&dpif_mutex);SHASH_FOR_EACH(node, &dpif_classes) {const struct registered_dpif_class *registered_class = node->data;sset_add(types, registered_class->dpif_class->type);}ovs_mutex_unlock(&dpif_mutex);
}

dp_enumerate_types列出所有支持的dpif_class，目前主要有dpif_netdev_class以及dpif_netlink_class两类

ofproto_type_run实际调用了ofproto_dpif_class的type_run函数，里面有调用了dpif_class->run，以及udpif_run
struct udpif代表了ofproto_dpif的upcall handler结构体，包含两部分，一是struct handler的数组，用于处理upcall请求，可以看做是一个线程池，二是revalidators的数组，这块我没有细看，应该是一种类似gc机制的线程池，用于回收过期的flow

/* An upcall handler for ofproto_dpif.** udpif keeps records of two kind of logically separate units:** upcall handling* ---------------**    - An array of 'struct handler's for upcall handling and flow*      installation.** flow revalidation* -----------------**    - Revalidation threads which read the datapath flow table and maintains*      them.*/
struct udpif {struct ovs_list list_node;         /* In all_udpifs list. */struct dpif *dpif;                 /* Datapath handle. */struct dpif_backer *backer;        /* Opaque dpif_backer pointer. */struct handler *handlers;          /* Upcall handlers. */size_t n_handlers;struct revalidator *revalidators;  /* Flow revalidators. */size_t n_revalidators;struct latch exit_latch;           /* Tells child threads to exit. *//* Revalidation. */struct seq *reval_seq;             /* Incremented to force revalidation. */bool reval_exit;                   /* Set by leader on 'exit_latch. */struct ovs_barrier reval_barrier;  /* Barrier used by revalidators. */struct dpif_flow_dump *dump;       /* DPIF flow dump state. */long long int dump_duration;       /* Duration of the last flow dump. */struct seq *dump_seq;              /* Increments each dump iteration. */atomic_bool enable_ufid;           /* If true, skip dumping flow attrs. *//* These variables provide a mechanism for the main thread to pause* all revalidation without having to completely shut the threads down.* 'pause_latch' is shared between the main thread and the lead* revalidator thread, so when it is desirable to halt revalidation, the* main thread will set the latch. 'pause' and 'pause_barrier' are shared* by revalidator threads. The lead revalidator will set 'pause' when it* observes the latch has been set, and this will cause all revalidator* threads to wait on 'pause_barrier' at the beginning of the next* revalidation round. */bool pause;                        /* Set by leader on 'pause_latch. */struct latch pause_latch;          /* Set to force revalidators pause. */struct ovs_barrier pause_barrier;  /* Barrier used to pause all *//* revalidators by main thread. *//* Following fields are accessed and modified by different threads. */atomic_uint flow_limit;            /* Datapath flow hard limit. *//* n_flows_mutex prevents multiple threads updating these concurrently. */atomic_uint n_flows;               /* Number of flows in the datapath. */atomic_llong n_flows_timestamp;    /* Last time n_flows was updated. */struct ovs_mutex n_flows_mutex;/* Following fields are accessed and modified only from the main thread. */struct unixctl_conn **conns;       /* Connections waiting on dump_seq. */uint64_t conn_seq;                 /* Corresponds to 'dump_seq' whenconns[n_conns-1] was stored. */size_t n_conns;                    /* Number of connections waiting. */
};

struct handler以及struct revalidator

/* A thread that reads upcalls from dpif, forwards each upcall's packet,* and possibly sets up a kernel flow as a cache. */
struct handler {struct udpif *udpif;               /* Parent udpif. */pthread_t thread;                  /* Thread ID. */uint32_t handler_id;               /* Handler id. */
};/* A thread that processes datapath flows, updates OpenFlow statistics, and* updates or removes them if necessary.** Revalidator threads operate in two phases: "dump" and "sweep". In between* each phase, all revalidators sync up so that all revalidator threads are* either in one phase or the other, but not a combination.**     During the dump phase, revalidators fetch flows from the datapath and*     attribute the statistics to OpenFlow rules. Each datapath flow has a*     corresponding ukey which caches the most recently seen statistics. If*     a flow needs to be deleted (for example, because it is unused over a*     period of time), revalidator threads may delete the flow during the*     dump phase. The datapath is not guaranteed to reliably dump all flows*     from the datapath, and there is no mapping between datapath flows to*     revalidators, so a particular flow may be handled by zero or more*     revalidators during a single dump phase. To avoid duplicate attribution*     of statistics, ukeys are never deleted during this phase.*     During the sweep phase, each revalidator takes ownership of a different*     slice of umaps and sweeps through all ukeys in those umaps to figure out*     whether they need to be deleted. During this phase, revalidators may*     fetch individual flows which were not dumped during the dump phase to*     validate them and attribute statistics.*/
struct revalidator {struct udpif *udpif;               /* Parent udpif. */pthread_t thread;                  /* Thread ID. */unsigned int id;                   /* ovsthread_id_self(). */
};

之后对每个bridge，调用ofproto_run

netdev_run

/* Performs periodic work needed by all the various kinds of netdevs.** If your program opens any netdevs, it must call this function within its* main poll loop. */
void
netdev_run(void)OVS_EXCLUDED(netdev_mutex)
{netdev_initialize();struct netdev_registered_class *rc;CMAP_FOR_EACH (rc, cmap_node, &netdev_classes) {if (rc->class->run) {rc->class->run(rc->class);}}
}

netdev_run会首先调用netdev_initialize，该函数用一个struct ovsthread_once代码块来管理，保证只会调用一次。其主要工作是注册了不同的netdev_class和tunnel_class

static void
netdev_initialize(void)OVS_EXCLUDED(netdev_mutex)
{static struct ovsthread_once once = OVSTHREAD_ONCE_INITIALIZER;if (ovsthread_once_start(&once)) {fatal_signal_add_hook(restore_all_flags, NULL, NULL, true);netdev_vport_patch_register();netdev_register_provider(&netdev_linux_class);netdev_register_provider(&netdev_internal_class);netdev_register_provider(&netdev_tap_class);netdev_vport_tunnel_register();ovsthread_once_done(&once);}
}

netdev_vport_patch_register注册了patch port这么一类的vport（patch port用于连接不同的bridge，本身没有dpif的任何属性），netdev_vport_tunnel_register则注册了多种tunnel类型，e.g.

    static const struct vport_class vport_classes[] = {TUNNEL_CLASS("geneve", "genev_sys", netdev_geneve_build_header,netdev_tnl_push_udp_header,netdev_geneve_pop_header),TUNNEL_CLASS("gre", "gre_sys", netdev_gre_build_header,netdev_gre_push_header,netdev_gre_pop_header),TUNNEL_CLASS("ipsec_gre", "gre_sys", NULL, NULL, NULL),TUNNEL_CLASS("vxlan", "vxlan_sys", netdev_vxlan_build_header,netdev_tnl_push_udp_header,netdev_vxlan_pop_header),TUNNEL_CLASS("lisp", "lisp_sys", NULL, NULL, NULL),TUNNEL_CLASS("stt", "stt_sys", NULL, NULL, NULL),};

除此之外，多个类型的netdev_class也被注册进来，包括dpdk的多个netdev_class

const struct netdev_class netdev_linux_class =NETDEV_LINUX_CLASS("system",netdev_linux_construct,netdev_linux_get_stats,netdev_linux_get_features,netdev_linux_get_status);const struct netdev_class netdev_tap_class =NETDEV_LINUX_CLASS("tap",netdev_linux_construct_tap,netdev_tap_get_stats,netdev_linux_get_features,netdev_linux_get_status);const struct netdev_class netdev_internal_class =NETDEV_LINUX_CLASS("internal",netdev_linux_construct,netdev_internal_get_stats,NULL,                  /* get_features */netdev_internal_get_status);static const struct netdev_class dpdk_class =NETDEV_DPDK_CLASS("dpdk",netdev_dpdk_construct,netdev_dpdk_destruct,netdev_dpdk_set_config,netdev_dpdk_set_tx_multiq,netdev_dpdk_eth_send,netdev_dpdk_get_carrier,netdev_dpdk_get_stats,netdev_dpdk_get_features,netdev_dpdk_get_status,netdev_dpdk_reconfigure,netdev_dpdk_rxq_recv);static const struct netdev_class dpdk_ring_class =NETDEV_DPDK_CLASS("dpdkr",netdev_dpdk_ring_construct,netdev_dpdk_destruct,netdev_dpdk_ring_set_config,netdev_dpdk_set_tx_multiq,netdev_dpdk_ring_send,netdev_dpdk_get_carrier,netdev_dpdk_get_stats,netdev_dpdk_get_features,netdev_dpdk_get_status,netdev_dpdk_reconfigure,netdev_dpdk_rxq_recv);static const struct netdev_class dpdk_vhost_class =NETDEV_DPDK_CLASS("dpdkvhostuser",netdev_dpdk_vhost_construct,netdev_dpdk_vhost_destruct,NULL,NULL,netdev_dpdk_vhost_send,netdev_dpdk_vhost_get_carrier,netdev_dpdk_vhost_get_stats,NULL,NULL,netdev_dpdk_vhost_reconfigure,netdev_dpdk_vhost_rxq_recv);

最后对于每一类注册的netdev_class，都会调用run函数，以netdev_linux_class为例（dpdk_class的run函数为空），主要是通过netlink sock来定期做一些更新的工作，不细说了

原文链接：https://blog.csdn.net/majieyue/article/details/52801322

OVS vswitchd启动（三十九）相关推荐

javaweb学习总结(三十九)——数据库连接池
javaweb学习总结(三十九)--数据库连接池一.应用程序直接获取数据库连接的缺点用户每次请求都需要向数据库获得链接,而数据库创建连接通常需要消耗相对较大的资源,创建时间也较长.假设网站一天10 ...
【正点原子STM32连载】第三十九章 DS18B20数字温度传感器实验摘自【正点原子】MiniPro STM32H750 开发指南_V1.1
1)实验平台:正点原子MiniPro H750开发板 2)平台购买地址:https://detail.tmall.com/item.htm?id=677017430560 3)全套实验源码+手册+视频 ...
孤傲苍狼只为成功找方法，不为失败找借口！ javaweb学习总结(三十九)——数据库连接池一、应用程序直接获取数据库连接的缺点　　用户每次请求都需要向数据库获得链接，而数据库创建连接通常需要
孤傲苍狼只为成功找方法,不为失败找借口! javaweb学习总结(三十九)--数据库连接池一.应用程序直接获取数据库连接的缺点用户每次请求都需要向数据库获得链接,而数据库创建连接通常需要消耗相对 ...
【正点原子MP157连载】第三十九章 LCD驱动实验-摘自【正点原子】STM32MP1嵌入式Linux驱动开发指南V1.7
1)实验平台:正点原子STM32MP157开发板 2)购买链接:https://item.taobao.com/item.htm?&id=629270721801 3)全套实验源码+手册+视频 ...
NeHe OpenGL第三十九课：物理模拟
NeHe OpenGL第三十九课:物理模拟物理模拟简介: 还记得高中的物理吧,直线运动,自由落体运动,弹簧.在这一课里,我们将创造这一切. 物理模拟介绍如果你很熟悉物理规律,并且想实现它,这篇 ...
Python编程基础：第三十九节面向对象编程Object Oriented Programming
第三十九节面向对象编程Object Oriented Programming 前言实践前言到目前为止我们都是函数式编程,也即将每一个功能块写为一个函数.其实还有一种更常用的编程方式被称为面向对 ...
三十九、Java集合中的HashSet和TreeSet
@Author:Runsen @Date:2020/6/6 作者介绍:Runsen目前大三下学期,专业化学工程与工艺,大学沉迷日语,Python, Java和一系列数据分析软件.导致翘课严重,专业排名 ...
WPF，Silverlight与XAML读书笔记第三十九 - 可视化效果之3D图形
原文:WPF,Silverlight与XAML读书笔记第三十九 - 可视化效果之3D图形说明:本系列基本上是<WPF揭秘>的读书笔记.在结构安排与文章内容上参照<WPF揭秘> ...
【零基础学Java】—List集合（三十九）
[零基础学Java]-List集合(三十九) java.util.list接口 extends Collection接口 list接口的特点: 1.有序的集合,存储元素和取出元素的顺序是一致的(存储1 ...
JavaScript学习（三十九）—对象中内容的操作
JavaScript学习(三十九)-对象中内容的操作一.对象中内容的操作:增.删.改.查 (一).增:给对象添加属性或者方法 1)方式1:对象名称.属性名=属性值: 2)方式2:对象名称['属性名' ...

OVS vswitchd启动（三十九）

ovs vswitchd的启动

bridge_run

netdev_run

OVS vswitchd启动（三十九）相关推荐

最新文章

热门文章